跨被试脑电表征的异质性与分布对齐需求

针对跨学科EEG情绪识别中个体差异显著和时频特征泛化不足的挑战，本文提出了一种多频时空注意力网络FA-TSception。该模型创新性地整合了多频率自适应机制和高效的通道注意力，构建了一个基于TSeption多尺度时空架构的三级处理框架。多频动态时间层通过参数化比例因子生成自适应卷积核组，以精确匹配Alpha、Beta、Gamma等情绪相关频带的时频特征；非对称空间层结合半球卷积核提取前额叶和时间区域的空间激活模式；集成了高效的信道注意力模块(ECA)，实现了多频特征的自适应校准。DEAP数据集上的跨学科实验表明，FA-TSception在唤醒和效价维度上的平均分类准确率分别达到62.73%和60.12%。与TSception相比，它提高了1.16%，仅增加了5.6%的模型参数计数。FA-TSception不仅提高了跨个体EEG情绪识别的准确性，而且通过引入有效的注意力机制，同时保持相对稳定的模型参数数量，增强了模型识别情绪相关特征的能力。

针对传统脑电情绪识别方法中存在的特征表征单维化、时序依赖建模不充分及任务间关联性被忽视等局限，本文提出一种面向多任务学习的时空混合注意力脑电情绪识别模型(MT-STCNN)。该方法首先在特征层面融合微分熵(DE)与功率谱密度(PSD)特征，从信号复杂度与频域能量分布两个维度全面表征情绪状态；在网络架构中引入轻量级Transformer编码器，强化对长时序脑电信号的全局依赖建模能力；进一步构建多任务协同学习框架，联合优化效价与唤醒度两项情感维度识别任务，通过特征共享机制与跨任务注意力模块实现任务间的知识迁移与互补，提升模型的识别效率。在DEAP数据集上的实验结果显示，本文方法在效价与唤醒度识别上的平均准确率分别达到98.26%和98.67%，性能显著优于当前主流模型，充分验证了所提方法在脑电情绪识别任务中的有效性与先进性。

基于运动想象的脑机接口(Brain Computer Interface, BCI)可以帮助残疾人控制机械手臂等外部设备，其中脑电信号解码是关键所在。但是不同个体间的脑电信号差异很大，使得传统的深度学习模型所采用的静态卷积很难自适应地提取脑电特征。为解决这个问题，本文提出了基于多维动态卷积的深度学习模型(Multidimensional Dynamic Convolution Net, MDconvnet)，该模型通过三层多维动态卷积来提取特征，并将提取的特征输入到全连接层来获取分类结果。其中多维动态卷积会依据输入的数据，生成卷积多维度的注意力权重，并将该权重与卷积参数相乘来动态地调节卷积参数，以便更好地挖掘数据时空特征。本文采用2023运动想象数据集RankA和数据集RankB对MDConvnet模型进行了测试，同时与多个经典的运动想象识别模型(FBCSP、EEGnet、EEGTCN、FBCnet、Tesecption、STASCNN、Deepconvnet和VIT)进行性能对比。结果显示MDConvnet模型在RankA和RankB数据集上的平均准确率分别为64.20%和67.04%，超过其他算法模型，展现出了MDConvnet模型在运动想象脑电识别任务上的优异性能，为残疾人通过脑机接口控制外部设备提供了有力的支持。

情感识别是人机交互中一个比较关键的问题，脑电作为人生理号中重要的组成部分，是识别人情感的关键因素。由于大脑中复杂的神经元活动会导致信号的非平稳行为，利用脑电进行情绪识别是一项具有挑战性的研究，并且多通道的分析是处理脑电信号时需要考虑的重要问题。本文提出了一种基于长短时间记忆网络与集成学习的多通道脑电情感识别模型，通过研究选取对情绪反馈响应较大的脑电通道，将不同通道的脑电数据分别放入长短时记忆网络中进行训练，再将各通道的训练模型通过集成学习策略进行优化，从而能整合各通道的信息进行情感分类。此研究在DEAP基准数据集上进行了情绪识别实验，我们在效价和唤醒两个情感维度的的情感识别结果均有明显的提升，准确率分别达到了87.6%，90.52%，验证了多通道集成方法的有效性。

针对运动想象(Motor Imagery, MI)脑电信号(EEG)普遍存在的信噪比低、非平稳性强以及传统卷积神经网络在特征提取过程中易产生大量时空冗余信息的问题，本文提出了一种融合多尺度局部特征选择与特征重构机制的深度学习解码模型。首先，该模型在浅层特征提取后引入多尺度局部特征选择模块，通过并行的多尺度深度卷积捕获不同感受野下的特征，并利用可学习的通道注意力权重实现特征的自适应加权，以增强特征的判别性。其次，为了进一步抑制任务无关的冗余信息，模型引入了空间与通道重构卷积(SCConv)特征优化模块，通过空间重构单元(SRU)和通道重构单元(CRU)对特征图进行压缩与重组，从而显著提升特征表达的有效性。在大型公开数据集OpenBMI上的实验结果表明，该模型在运动想象任务中的平均准确率达到72.95%，优于EEGNet、Conformer等主流对比方法。消融实验进一步证实了多尺度特征选择模块与SCConv特征优化模块在提升模型鲁棒性和解码性能方面的关键作用。

运动想象是一种认知神经科学领域的概念，指的是在不实际运动的情况下，通过想象运动来激活大脑相应区域的神经元。传统的CNN在处理EEG信号时存在劣势，因为EEG信号是一种时间序列数据，而CNN并不擅长处理这种类型的数据，导致无法充分挖掘时间相关性和特征信息，影响了模型的性能和准确性。为了解决这一问题，本文使用动态图卷积和时间卷积来处理EEG数据，该方法能够有效地捕捉信号之间的时间依赖关系和动态变化，从而提高了模型在处理EEG信号时的性能和准确性。动态图卷积的优势在于能够更好地适应时间序列数据的特点，提高了模型在提取特征和预测方面的效果，有效解决了传统CNN在处理EEG信号时的劣势，为脑机接口技术等领域的发展带来了新的可能性。该方法主要过程如下：首先，EEG信号被输入到卷积滤波器进行处理，过滤成八个子频带后，分别输入到八个动态图卷积神经网络(DGCNN)中。最后，这些网络被串联起来，输入到一个时域卷积网络(TCN)中进行特征提取。在公开数据集上，DGCNN模型的平均分类准确率(82.5 ± 4.3%)优于传统的CNN模型(68.9 ± 3.6%)。

疲劳驾驶是交通事故的重要诱因，其检测对于保障交通安全至关重要。脑电图(EEG)信号因能反映大脑活动状态而被广泛用于疲劳驾驶检测，但现有深度学习方法常面临对预处理的依赖或对EEG信号时空信息的处理不足等挑战。本文提出了一种基于脑电图信号的时空融合疲劳驾驶检测方法——KCL-STN (KAN-CNN-LSTM Spatio-Temporal information Network)。该方法巧妙地结合了卷积神经网络和长短期记忆网络，分别从原始脑电信号中提取空间和时间特征，并进行有效融合，实现了端到端的疲劳驾驶检测。针对脑电数据稀缺问题，本文还提出了一种脑电信号滑动窗口增强算法，以增加样本数量并提高模型训练的稳定性。在公开数据集上的实验结果表明，KCL-STN在分类准确度、召回率和精确率等指标上均优于多种现有方法，准确率达到86.05%。消融实验证实了关键组件KAN线性层和滑动窗口数据增强方法的有效性。跨被试实验也证明了模型良好的泛化性能和鲁棒性。研究结果表明，KCL-STN能够有效地从原始脑电信号中提取疲劳相关特征，是鲁棒且高性能的疲劳驾驶检测方法。

EEG emotion recognition has been hampered by the clear individual differences in the electroencephalogram (EEG). Nowadays, domain adaptation is a good way to deal with this issue because it aligns the distribution of data across subjects. However, the performance for EEG emotion recognition is limited by the existing research, which mainly focuses on the global alignment between the source domain and the target domain and ignores much fine-grained information. In this study, we propose a method called Graph-based Unsupervised Subdomain Adaptation (Gusa), which simultaneously aligns the distribution between the source and target domains in a fine-grained way from both the channel and emotion subdomains. Gusa employs three modules, such as the Node-wise Domain Constraints Module to align each EEG channel and obtain a domain-variant representation, the Class-level Distribution Constraints Module, and the Emotion-wise Domain Constraints Module, to collect more fine-grained information, create more discriminative representations for each emotion, and lessen the impact of noisy emotion labels. The studies on the SEED, SEED-IV, and MPED datasets demonstrate that Gusa significantly improves the ability of EEG to recognize emotions and can extract more granular and discriminative representations for EEG.

Currently, most models rarely consider the negative transfer problem in the research field of cross-subject EEG emotion recognition. To solve this problem, this paper proposes a semi-supervised domain adaptive algorithm based on few labeled samples of target subject, which called multi-domain geodesic flow kernel dynamic distribution alignment (MGFKD). It consists of three modules: 1) GFK common feature extractor: projects the feature distribution of source and target subjects to the Grassmann manifold space, and obtains the latent common features of the two feature distributions through GFK method. 2) Source domain selector: obtains pseudo-labels of the target subject through weak classifier, finds "golden source subjects" by using few known labels of target subjects. 3) Label corrector: uses a dynamic distribution balance strategy to correct the pseudo-labels of the target subject. We conducted comparison experiments on the SEED and SEED-IV datasets, and the results show that MGFKD outperforms unsupervised and semi-supervised domain adaptation algorithms, achieving an average accuracy of 87.51±7.68% and 68.79±8.25% on the SEED and SEED-IV datasets with only one labeled sample per video for target subject. Especially when the number of source domains is set as 6 and the number of known labels is set as 5, the accuracy increase to 90.20±7.57% and 69.99±7.38%, respectively. The above results prove that our proposed algorithm can efficiently and quickly improve the cross-subject EEG emotion classification performance. Since it only need a small number of labeled samples of new subjects, making it has strong application value in future EEG-based emotion recognition applications.

Domainadaptation has proven effective for suppressing the inter-subject variability problem in cross-subject EEG classification tasks in which labeled data is available for source subjects while only unlabeled data is provided for target subjects. Existing domain adaptation methods typically reduced the distribution discrepancy between source and target domains by directly utilizing source domain samples or features. To safeguard the privacy of source domain data, we propose to construct a Proxy Domain by simultaneously considering the prediction Consistency and Confidence (PDCC) of locally trained source models on target EEG samples, serving as the substitute to the source domain. The framework commences with the augmentation and alignment of the source domain data to enhance feature generalizability, after which source models are trained independently on each source subject’s data in a decentralized manner. Knowledge transfer from source to target domains is achieved exclusively through accessing to the source domain model, enabling the PDCC-based proxy domain construction that encapsulates the source knowledge. Finally, domain adaptation is performed using the proxy domain and target domain. As a result, PDCC eliminates the need to access source domain data while effectively leveraging source knowledge. Experimental results on four benchmark EEG datasets demonstrate that PDCC consistently outperforms eleven existing methods, including several advanced transfer learning and source-free methods. Especially, the effectiveness of the proxy domain is extensively investigated.

The collection of substantial amounts of electroencephalogram (EEG) data is typically time-consuming and labor-intensive, which adversely impacts the development of decoding models with strong generalizability, particularly when the available data is limited. Utilizing sufficient EEG data from other subjects to aid in modeling the target subject presents a potential solution, commonly referred to as domain adaptation. Most current domain adaptation techniques for EEG decoding primarily focus on learning shared feature representations through domain alignment strategies. Since the domain shift cannot be completely removed, target EEG samples located near the edge of clusters are also susceptible to misclassification. We propose a novel adaptive deep feature representation (ADFR) framework to improve the cross-subject EEG classification performance through learning transferable EEG feature representations. Specifically, we first minimize the distribution discrepancy between the source and target domains by employing maximum mean discrepancy (MMD) regularization, which aids in learning the shared feature representations. We then utilize the instance-based discriminative feature learning (IDFL) regularization to make the learned feature representations more discriminative. Finally, the entropy minimization (EM) regularization is further integrated to adjust the classifier to pass through the low-density region between clusters. The synergistic learning between above regularizations during the training process enhances EEG decoding performance across subjects. The effectiveness of the ADFR framework was evaluated on two public motor imagery (MI)-based EEG datasets: BCI Competition III dataset 4a and BCI Competition IV dataset 2a. In terms of average accuracy, ADFR achieved improvements of 3.0% and 2.1%, respectively, over the state-of-the-art methods on these datasets. The promising results highlight the effectiveness of the ADFR algorithm for EEG decoding and show its potential for practical applications.

Emotion recognition based on electroencephalography (EEG) holds significant promise for affective brain-computer interfaces (aBCIs). However, its practical deployment faces challenges due to the variability within inter-subject and the scarcity of labeled data in target domains. To overcome these limitations, we propose SDC-Net, a novel Semantic-Dynamic Consistency domain adaptation network for fully label-free cross-subject EEG emotion recognition. First, we introduce a Same-Subject Same-Trial Mixup strategy that generates augmented samples through intra-trial interpolation, enhancing data diversity while explicitly preserving individual identity to mitigate label ambiguity. Second, we construct a dynamic distribution alignment module within the Reproducing Kernel Hilbert Space (RKHS), jointly aligning marginal and conditional distributions through multi-objective kernel mean embedding, and leveraging a confidence-aware pseudo-labeling strategy to ensure stable adaptation. Third, we propose a dual-domain similarity consistency learning mechanism that enforces cross-domain structural constraints based on latent pairwise similarities, facilitating semantic boundary learning without reliance on temporal synchronization or label priors. To validate the effectiveness and robustness of the proposed SDC-Net, extensive experiments are conducted on three widely used EEG benchmark datasets: SEED, SEED-IV, and FACED. Comparative results against existing unsupervised domain adaptation methods demonstrate that SDC-Net achieves state-of-the-art performance in emotion recognition under both cross-subject and cross-session conditions. This advancement significantly improves the accuracy and generalization capability of emotion decoding, laying a solid foundation for real-world applications of personalized aBCIs. The source code is available at: https://github.com/XuanSuTrum/SDC-Net.

Electroencephalography (EEG) has seen rapidly growing applications in healthcare and brain-computer interfaces, yet cross-subject generalization remains a critical challenge. Unsupervised Domain Adaptation (UDA) techniques are widely employed to mitigate domain shift, but the strong constraints introduced during adaptation often lead to feature space collapse and loss of information. Moreover, relying on empirical risk minimization with cross-entropy (CE) loss fails to effectively prevent such collapse. To address this issue, we propose a novel regularizer, Dynamic Log-Determinant Gating Loss (Dyn-LogDet). This strategy introduces a history-adaptive gating mechanism based on the log-determinant of the feature Gram matrix, such that the penalty is applied only when the current feature diversity drops below a historical baseline. This ultimately contributing to better generalization performance. We validate Dyn-LogDet on two representative categories of UDA approaches: (i) explicit distribution alignment, represented by the maximum mean discrepancy (MMD)-based regularization method, and (ii) adversarial distribution alignment, represented by adversarial learning approaches(ADV), and conduct experiments on the SEED and SEED-IV datasets. In both cases, Dyn-LogDet yields consistent and improvements in performance, demonstrating its efficacy in enhancing cross-subject generalization.

Objective. Domain adaptation (DA) has achieved remarkable performance in cross-subject electroencephalogram (EEG) decoding by mitigating the inter-subject data distribution discrepancies. However, when exploring the feature alignment subspace and performing self-supervised pseudo-labeling in an iterative way, two difficulties are often encountered: one is that unreliable target labeling results inevitably mislead the domain-free feature learning process in the early stage and the other is that the contribution of source and target samples should be balanced in the later stage. Approach. To address both issues, this paper proposes prototype-based progressive confident target sample labeling (P2CSL) method to use subspace class prototypes to assist in labeling target samples under the unified framework of domain-invariant EEG feature learning and the self-supervised target sample labeling, and progressively incorporate confident target samples into DA model fitting. The underlying rationality is that early-stage pseudo-labels from unconverged models are prone to error propagation, requiring auxiliary mechanisms to ensure their reliability and stabilize training. With the gradual alignment of cross-subject features, the estimated pseudo-label information of target domain will be more reliable, meaning that more target samples should be involved in model training. Main results. Experiments on emotion recognition and inner speech decoding demonstrate the competitive performance of P2CSL in cross-subject EEG classification in comparison with SOTA methods. Significance. Our study indicates the effectiveness of jointly considering the reliability of target samples and their contribution to model training in the context of DA. In addition, some fine-grained results including the sample confidence allocation strategy, the DA effects, and the dynamic model optimization process are provided to further illustrate the model execution details.

Objective. Many subject-dependent methods were proposed for electroencephalogram (EEG) classification in rapid serial visual presentation (RSVP) task, which required a large amount of data from new subject and were time-consuming to calibrate system. Cross-subject classification can realize calibration reduction or zero calibration. However, cross-subject classification in RSVP task is still a challenge. Approach. This study proposed a multi-source domain adaptation based tempo-spatial convolution (MDA-TSC) network for cross-subject RSVP classification. The proposed network consisted of three modules. First, the common feature extraction with multi-scale tempo-spatial convolution was constructed to extract domain-invariant features across all subjects, which could improve generalization of the network. Second, the multi-branch domain-specific feature extraction and alignment was conducted to extract and align domain-specific feature distributions of source and target domains in pairs, which could consider feature distribution differences among source domains. Third, the domain-specific classifier was exploited to optimize the network through loss functions and obtain prediction for the target domain. Main results. The proposed network was evaluated on the benchmark RSVP dataset, and the cross-subject classification results showed that the proposed MDA-TSC network outperformed the reference methods. Moreover, the effectiveness of the MDA-TSC network was verified through both ablation studies and visualization. Significance. The proposed network could effectively improve cross-subject classification performance in RSVP task, and was helpful to reduce system calibration time.

Most emotion recognition systems still present limited applicability to new users due to the intersubject variability of electroencephalogram (EEG) signals. Although domain adaptation methods have been adopted to tackle this problem, most methodologies deal with unlabeled data from a target subject. However, a few labeled samples from a target subject could also be included to boost cross-subject emotion recognition. In this article, we present a semisupervised domain adaptation (SSDA) framework to align the joint distributions of subjects, assuming that fine-grained structures must be aligned to perform a greater knowledge transfer. To achieve this, the proposed framework performs a multisource alignment of features at the subject level, while predictions are aligned over the global feature space. To support joint distribution alignment, interclass separation and consistent predictions are ensured on the target subject. We perform experiments using two public benchmark datasets, SEED and SEED-IV, with two different sampling strategies to incorporate a few labeled samples from a target subject. Our proposal achieves an average accuracy of 93.55% and 87.96% on SEED and SEED-IV, using three labeled target samples of each class. Moreover, we obtained an average accuracy of 91.79% and 85.45% on SEED and SEED-IV by incorporating ten labeled samples from the first EEG trial of each class.

No abstract available

Multi-modal emotion recognition has attracted much attention in human-computer interaction, because it provides complementary information for the recognition model. However, the distribution drift among subjects and the heterogeneity of different modalities pose challenges to multi-modal emotion recognition, thereby limiting its practical application. Most of the current multi-modal emotion recognition methods are difficult to suppress above uncertainties in fusion. In this paper, we propose a cross-subject multi-modal emotion recognition framework, which jointly learns subject-independent representation and common feature between EEG and eye movements. First, we design the dynamic adversarial domain adaptation for cross-subject distribution alignment, dynamically selecting source domains in training. Second, we simultaneously capture intra-modal and inter-modal emotion-related features by both self-attention and cross-attention mechanisms, thus obtaining the robust and complementary representation of emotional information. Then, two contrastive loss functions are imposed on above network to further reduce inter-modal heterogeneity, and mine higher-order semantic similarity between synchronously collected multi-modal data. Finally, we used the output of the softmax layer as the predicted value. The experimental results on several multi-modal emotion datasets with EEG and eye movements demonstrate that our method is significantly superior to the state-of-the-art emotion recognition approaches.

Personalised music-based interventions offer a powerful means of supporting motor rehabilitation by dynamically tailoring auditory stimuli to provide external timekeeping cues, modulate affective states, and stabilise gait patterns. Generalisable Brain-Computer Interfaces (BCIs) thus hold promise for adapting these interventions across individuals. However, inter-subject variability in EEG signals, further compounded by movement-induced artefacts and motor planning differences, hinders the generalisability of BCIs and results in lengthy calibration processes. We propose Individual Tangent Space Alignment (ITSA), a novel pre-alignment strategy incorporating subject-specific recentering, distribution matching, and supervised rotational alignment to enhance cross-subject generalisation. Our hybrid architecture fuses Regularised Common Spatial Patterns (RCSP) with Riemannian geometry in parallel and sequential configurations, improving class separability while maintaining the geometric structure of covariance matrices for robust statistical computation. Using leave-one-subject-out cross-validation, `ITSA'demonstrates significant performance improvements across subjects and conditions. The parallel fusion approach shows the greatest enhancement over its sequential counterpart, with robust performance maintained across varying data conditions and electrode configurations. The code will be made publicly available at the time of publication.

No abstract available

EEG-based emotion recognition has advanced rapidly due to its objectivity and reliability, but individual differences present challenges: subject-specific models underperform on new subjects, and general models lack accuracy. While domain adaptation (DA) algorithms reduce distribution differences between source and target EEG domains, single-source methods struggle with knowledge transfer, and multi-source methods neglect source domain differences. To address this, we propose a collaborative learning and dynamic distributed adaptation algorithm (CL-DDA) for EEG emotion recognition. By dividing the model into two sub-networks, collaborative learning improves generalization. Additionally, local subdomain alignment addresses inter-subject emotion differences, while global domain alignment minimizes marginal distribution disparities. Our model achieved 90.08% and 77.55% accuracy on SEED and SEED-IV datasets, respectively, in cross-subject emotion recognition.

Emotion decoding using Electroencephalography (EEG)-based affective brain-computer interfaces (aBCIs) is crucial for affective computing but is hindered by EEG’s non-stationarity, individual variability, and the high cost of large-scale labeled data. Deep learning-based approaches, while effective, require substantial computational resources and large datasets, limiting their practicality. To address these challenges, we propose Manifold-based Domain Adaptation with Dynamic Distribution (M3D), a lightweight non-deep transfer learning framework. M3D includes four main modules: manifold feature transformation, dynamic distribution alignment, classifier learning, and ensemble learning. The data undergoes a transformation onto an optimal Grassmann manifold space, enabling dynamic alignment of the source and target domains. This process prioritizes both marginal and conditional distributions according to their significance, ensuring enhanced adaptation efficiency across various types of data. In the classifier learning, the principle of structural risk minimization is integrated to develop robust classification models. This is complemented by dynamic distribution alignment, which refines the classifier iteratively. Additionally, the ensemble learning module aggregates the classifiers obtained at different stages of the optimization process, which leverages the diversity of the classifiers to enhance the overall prediction accuracy. The proposed M3D framework is evaluated on three benchmark EEG emotion recognition datasets using two validation protocols (cross-subject single-session and cross-subject cross-session), as well as on a clinical EEG dataset of Major Depressive Disorder (MDD). Experimental results demonstrate that M3D outperforms traditional non-deep learning methods, achieving an average improvement of 6.67%, while achieving deep learning-comparable performance with significantly lower data and computational requirements. These findings highlight the potential of M3D to enhance the practicality and applicability of aBCIs in real-world scenarios.

Multimodal emotion recognition from electroencephalogram (EEG) and eye movement signals has shown to be a promising approach to provide more discriminative information about human emotional states. However, most current works rely on a subject-dependent approach, limiting their applicability to new users. Recently, some studies have explored multimodal domain adaptation to address the mentioned issue by transferring information from known subjects to new ones. Unfortunately, existing methods are still exposed to negative transfer as a suboptimal distribution alignment is performed between subjects, while irrelevant information is not discarded. In this article, we present a multimodal and multisource domain adaptation (MMDA) method, which adopts the following three strategies: 1) marginal and conditional distribution alignments must be performed between each known subject and a new one; 2) relevant distribution alignments must be prioritized to avoid a negative transfer; and 3) modality fusion results should be improved by extracting more discriminative features from EEG signals and selecting relevant features across modalities. Our proposed method was evaluated with leave-one-subject-out cross validation on four public datasets: SEED, SEED-GER, SEED-IV, and SEED-V. Experimental results show that our proposal outperforms state-of-the-art results for each dataset when subject data from different sessions are combined into a single dataset. Moreover, MMDA exceeds the state of the art in 8 out of 11 different sessions when each session is evaluated.

No abstract available

Individual fluctuations and temporal variability of electroencephalogram (EEG) data pose challenges in precisely identifying emotions. Although a model may perform well with data specific to a certain subject or session, the fluctuations in EEG data can significantly impair the model's performance on a different subject or session. To tackle this problem, current approaches synchronize the original and new subject or session feature distributions. Directly matching EEG data across individuals or sessions may undermine the inherent distinguishability due to the heterogeneity in data distribution. Instead of direct alignment, this work utilizes multisource structural deep clustering to identify the inherent structural knowledge of the target itself and regularize it through the distribution of source labels. Furthermore, the method was implemented on the intermediate output utilizing high-confidence features to improve pattern identification in the latent feature space. This led to more distinct differentiations across subdomains with varying labels. Comparative analyses were performed with state of-the-art (SOTA) models on SEED and SEED-IV datasets. The model proposed outperformed other baseline models, reaching an average accuracy of 90.69%/95.05% in a cross-subject/cross session experiment on SEED and 74.35%/78.56% in SEED-IV. This research provides a novel approach to align EEG features without the need for direct distance calculation.

Motor imagery electroencephalogram (MI-EEG) plays a crucial role in developing brain-computer interfaces (BCIs) that enable natural interaction and autonomous control. Nevertheless, the performance of deep learning models is often hindered by inter-subject variability, which compromises their generalization across different domains. To address this issue, we introduce RGFTSLANet-a novel framework that combines a TSLANet module for signal denoising and temporal pattern extraction, a spatio-temporal convolution component for capturing higher-order features, and a multi-scale Riemannian geometry module to learn subject-invariant distribution characteristics. These complementary features are integrated to improve classification performance. Furthermore, to mitigate inter-subject distribution shifts, RGFTSLANet employs Maximum Mean Discrepancy (MMD) loss for domain adaptation, promoting alignment between source and target feature distributions. Experimental results on the four-class BCI Competition IV 2a dataset demonstrate the model's strong performance, achieving 71.39% accuracy in cross-subject evaluation and 80.71% in single-subject scenarios. Additional ablation and visualization studies confirm RGFTSLANet's capability to reduce cross-subject feature discrepancies and enhance the decoding of MI-EEG signals.

Auditory spatial attention detection (ASAD) seeks to determine which speaker in a surround sound field a listener is focusing on based on the one's brain biosignals. Although existing studies have achieved ASAD from a single-trial electroencephalogram (EEG), the huge inter-subject variability makes them generally perform poorly in cross-subject scenarios. Besides, most ASAD methods do not take full advantage of topological relationships between EEG channels, which are crucial for high-quality ASAD. Recently, some advanced studies have introduced graph-based brain topology modeling into ASAD, but how to calculate edge weights in a graph to better capture actual brain connectivity is worthy of further investigation. To address these issues, we propose a new ASAD method in this paper. First, we model a multi-channel EEG segment as a graph, where differential entropy serves as the node feature, and a static adjacency matrix is generated based on inter-channel mutual information to quantify brain functional connectivity. Then, different subjects' EEG graphs are encoded into a shared embedding space through a total variation graph neural network. Meanwhile, feature distribution alignment based on multi-kernel maximum mean discrepancy is adopted to learn subject-invariant patterns. Note that we align EEG embeddings of different subjects to reference distributions rather than align them to each other for the purpose of privacy preservation. A series of experiments on open datasets demonstrate that the proposed model outperforms state-of-the-art ASAD models in cross-subject scenarios with relatively low computational complexity, and feature distribution alignment improves the generalizability of the proposed model to a new subject.

Regarding cognitive workload recognition (CWR), electroencephalography (EEG) signals are nonstationary across time and vary from different subjects, thus hindering the cross-subject recognition performance. Although subject calibration or collecting massive training data may ease the above problem, it is generally time-consuming and expensive. In this article, we propose a deep domain adaptation (DDA) scheme for EEG-based cross-subject CWR, using the knowledge from the existing subjects (source domain) to improve the recognition performance of a new subject (target domain). Precisely, the proposed DDA method composes four modules, EEG features extractor, label classifier, feature distribution alignment, and domain discriminator. The model starts with the EEG feature extractor to learn the shallow feature representation for both domains. The label classifier further learns the deep representation and trains the classifier supervised. Finally, the feature distribution alignment matches the shallow feature distribution discrepancy, and the domain discriminator matches the deep distribution discrepancy by the adversarial training with the feature extractor. It not only learns domain-invariant features but also achieves robust domain-adaptive cross-subject recognition results in an end-to-end training framework. We conduct experiments to classify low and high workloads on a self-designed EEG dataset and one public EEG dataset. Experimental results demonstrate that our DDA scheme significantly outperforms the baselines and other state-of-the-art methods with improvements of 2%–9% in terms of recognition accuracy.

Decoding motor imagery (MI) electroencephalogram (EEG) signals in the brain–computer interface (BCI) can assist patients in accelerating motor function recovery. To realize the implementation of plug-and-play functionality for MI-BCI applications, cross-subject models are employed to alleviate time-consuming calibration and avoid additional model training for target subjects by utilizing EEG data from source subjects. However, the diversity in data distribution among subjects limits the model’s robustness. In this study, we investigate a cross-subject MI-EEG decoding model with domain generalization based on a deep learning neural network that extracts domain-invariant features from source subjects. Firstly, a knowledge distillation framework is adopted to obtain the internally invariant representations based on spectral features fusion. Then, the correlation alignment approach aligns mutually invariant representations between each pair of sub-source domains. In addition, we use distance regularization on two kinds of invariant features to enhance generalizable information. To assess the effectiveness of our approach, experiments are conducted on the BCI Competition IV 2a and the Korean University dataset. The results demonstrate that the proposed model achieves 8.93% and 4.4% accuracy improvements on two datasets, respectively, compared with current state-of-the-art models, confirming that the proposed approach can effectively extract invariant features from source subjects and generalize to the unseen target distribution, hence paving the way for effective implementation of the plug-and-play functionality in MI-BCI applications.

Decoding motor imagery (MI) electroencephalogram (EEG) signals, a key non-invasive brain-computer interface (BCI) paradigm for controlling external systems, has been significantly advanced by deep learning. However, MI-EEG decoding remains challenging due to substantial inter-subject variability and limited labeled target data, which necessitate costly calibration for new users. Many existing multi-source domain adaptation (MSDA) methods indiscriminately incorporate all available source domains, disregarding the large inter-subject differences in EEG signals, which leads to negative transfer and excessive computational costs. Moreover, while many approaches focus on feature distribution alignment, they often neglect the explicit dependence between features and decision-level outputs, limiting their ability to preserve discriminative structures. To address these gaps, we propose a novel MSDA framework that leverages a pretrained large Brain Foundation Model (BFM) for dynamic and informed source subject selection, ensuring only relevant sources contribute to adaptation. Furthermore, we employ Cauchy-Schwarz (CS) and Conditional CS (CCS) divergences to jointly perform feature-level and decision-level alignment, enhancing domain invariance while maintaining class discriminability. Extensive evaluations on two benchmark MI-EEG datasets demonstrate that our framework outperforms a broad range of state-of-the-art baselines. Additional experiments with a large source pool validate the scalability and efficiency of BFM-guided selection, which significantly reduces training time without sacrificing performance.

Automatic sleep stage classification is an effective technology compared to conventional artificial visual inspection in the field of sleep staging. Numerous algorithms based on machine learning and deep learning on single-channel electroencephalogram (EEG) have been proposed in recent years, however, category imbalance and cross-subject discrepancy are still the main factors restricting the accuracy of existing methods. This study proposed an innovative end-to-end neural network to solve these problems, specifically, four data augmentation methods were designed to eliminate category imbalance, and domain adaptation modules were designed for the alignment of marginal distribution, conditional distribution, and channel and spatial level distribution of feature maps, as well as the capture of transferable regions on the feature maps using a transfer attention mechanism. We conducted experiments on two publicly available datasets (Sleep-EDF Database Expanded, 2013 and 2018 version), Cohen's kappa coefficient (k) of 0.77 (Fpz-Cz) and 0.73 (Pz-Oz) were realized on the Sleep-EDF-2013 dataset, and a k of 0.75 (Fpz-Cz) and 0.68 (Pz-Oz) were realized on the Sleep-EDF-2018 dataset. An experiment was also conducted on the dataset drawn from the 2018 Physionet challenge, which containing people with sleep disorders, and a performance improvement was still found. Our comparative experiments with similar studies showed that our model was superior to most other studies, indicating our proposed EEG data augmentation and domain adaptation based cross-subject discrepancy alleviation approach is effective to improve the performance of automatic sleep staging.

Bioinformatics theoretical-based methods have attracted great attention in rehabilitation-assisted motor imagery (MI) brain-computer interface (BCI) using EEG, and gained promising results to deal with cross-subject MI-EEG classification. However, most of the existing methods are either suffered from outliers during feature learning across subjects or inefficient in computing discriminative features. Thus, they may not be able to obtain an optimal feature representation across subjects, resulting in classification performance deterioration. This paper proposes a dual regularization-based MI-EEG feature learning framework, in which feature weighted common spatial pattern, joint probability distribution alignment, and minimum Mahalanobis distance are taken into account, thus facilitating cross-subject classification. Specially, we adopt the multi-source to single target domain adaptation strategy, which is more suitable for real-world MI-BCI scenarios. Empirical studies on three benchmark MI-EEG datasets reveal the effectiveness and efficiency of the proposed method, which achieves a great performance improvement with less time consumption.

In cross-subject electroencephalography (EEG) motor imagery decoding tasks, significant physiological differences among individuals pose substantial challenges. Although Gaussian-based softmax Deep Domain Adaptation (DDA) methods have achieved considerable progress, they remain highly dependent on target domain data, which is inconsistent with real-world scenarios where target domain data may be inaccessible or extremely limited. Moreover, existing DDA methods primarily achieve feature alignment by minimizing distribution discrepancies between the source and target domains. However, given the pronounced physiological variability across individuals, simple distribution matching strategies often fail to effectively mitigate domain shift, thereby limiting generalization performance. To address these challenges, this study proposes an improved Gaussian-based softmax Deep Domain Generalization (Exp-G-softmax DDG) framework, which aims to overcome the limitations of traditional DDG methods in handling inter-class differences and cross-domain distribution shifts. By introducing multi-source domain joint training and an enhanced G-softmax function, the proposed method effectively resolves the dynamic balance between intra-class distance and inter-class distance. The Exp-G-softmax DDG mechanism integrates class center information, thereby enhancing model robustness and improving its ability to learn discriminative feature representations, ultimately leading to superior classification performance. Experimental results demonstrate that the proposed method achieves classification performance comparable to that of DDA on three publicly available real-world EEG datasets, providing a novel solution for cross-subject motor imagery decoding. The source code is available at: https://github.com/dawin2015/G-softmax-DDG .

Objective. Electroencephalogram (EEG) signals are promising biometrics owning to their invisibility, adapting to the application scenarios with high-security requirements. However, It is challenging to explore EEG identity features without the interference of device and state differences of the subject across sessions. Existing methods treat training sessions as a single domain, affected by the different data distribution among sessions. Although most multi-source unsupervised domain adaptation (MUDA) methods bridge the domain gap between multiple source and target domains individually, relationships among the domain-invariant features of each distribution alignment are neglected. Approach. In this paper, we propose a MUDA method, Tensorized Spatial-Frequency Attention Network (TSFAN), to assist the performance of the target domain for EEG-based biometric recognition. Specifically, significant relationships of domain-invariant features are modeled via a tensorized attention mechanism. It jointly incorporates appropriate common spatial-frequency representations of pairwise source and target but also cross-source domains, without the effect of distribution discrepancy among source domains. Additionally, considering the curse of dimensionality, our TSFAN is approximately represented in Tucker format. Benefiting the low-rank Tucker Network, the TSFAN can scale linearly in the number of domains, providing us the great flexibility to extend TSFAN to the case associated with an arbitrary number of sessions. Main results. Extensive experiments on the representative benchmarks demonstrate the effectiveness of TSFAN in EEG-based biometric recognition, outperforming state-of-the-art approaches, as verified by cross-session validation. Significance. The proposed TSFAN aims to investigate the presence of consistent EEG identity features across sessions. It is achieved by utilizing a novel tensorized attention mechanism that collaborates intra-source transferable information with inter-source interactions, while remaining unaffected by domain shifts in multiple source domains. Furthermore, the electrode selection shows that EEG-based identity features across sessions are distributed across brain regions, and 20 electrodes based on 10–20 standard system are able to extract stable identity information.

Large-scale foundation models for EEG signals offer a promising path to generalizable brain-computer interface (BCI) applications, but they often suffer from misalignment between pretraining objectives and downstream tasks, as well as significant cross-subject distribution shifts. This paper addresses these challenges by introducing a two-stage alignment strategy that bridges the gap between generic pretraining and specific EEG decoding tasks. First, we propose NeuroTTT: a domain-specific self-supervised fine-tuning paradigm that augments the foundation model with task-relevant self-supervised objectives, aligning latent representations to important spectral, spatial, and temporal EEG features without requiring additional labeled data. Second, we incorporate test-time training (TTT) at inference, we perform (i) self-supervised test-time training on individual unlabeled test samples and (ii) prediction entropy minimization (Tent), which updates only normalization statistics to continually calibrate the model to each new input on the fly. Our approach, which, to our knowledge, is the first to unify domain-tuned self-supervision with test-time training in large-scale EEG foundation models, yields substantially improved robustness and accuracy across diverse BCI tasks (imagined speech, stress detection, motor imagery). Using CBraMod and LaBraM as backbones, our method pushes their performance to a markedly higher level. Results on three diverse tasks demonstrate that the proposed alignment strategy achieves state-of-the-art performance, outperforming conventional fine-tuning and adaptation methods. Our code is available at https://github.com/wsl2000/NeuroTTT.

Task-specific pre-training is essential when task representations diverge from generic pre-training features. Existing task-general pre-training EEG models struggle with complex tasks like emotion recognition due to mismatches between task-specific features and broad pre-training approaches. This work aims to develop a task-specific multi-dataset joint pre-training framework for cross-dataset emotion recognition, tackling problems of large inter-dataset distribution shifts, inconsistent emotion category definitions, and substantial inter-subject variability. We introduce a cross-dataset covariance alignment loss to align second-order statistical properties across datasets, enabling robust generalization without the need for extensive labels or per-subject calibration. To capture the long-term dependency and complex dynamics of EEG, we propose a hybrid encoder combining a Mamba-like linear attention channel encoder and a spatiotemporal dynamics model. Our method outperforms state-of-the-art large-scale EEG models by an average of 4.57% in AUROC for few-shot emotion recognition and 11.92% in accuracy for zero-shot generalization to a new dataset. Performance scales with the increase of datasets used in pre-training. Multi-dataset joint pre-training achieves a performance gain of 8.55% over single-dataset training. This work provides a scalable framework for task-specific pre-training and highlights its benefit in generalizable affective computing. Our code is available at https://github.com/ncclab-sustech/mdJPT_nips2025.

Individual differences and nonstationary characteristics are prominent in electroencephalography (EEG) signals. Therefore, aligning the source and target domain data becomes essential in cross-subject and cross-session classification tasks. Although many adversarial adaptation networks can achieve distribution alignment through domain-level adaptation, they tend to disregard the multimodal structure inherent in the data. In this article, we present a concise and effective adversarial paradigm for EEG emotion recognition. This approach fully utilizes the label structure information of source domain data to reuse the binary discriminator as a class-informed discriminator instead of introducing additional modules, which not only realizes domain confusion but also ensures that mode information is retained in the process of confusion to avoid mode collapse. To evaluate our method, a systematic experimental study was conducted on the public datasets SEED and SEED-IV. The average accuracy of cross-subject and cross-session scenarios achieved 90.21%, 95.47% on SEED, and 77.50%, 77.54% on SEED-IV, respectively. Compared to the existing domain adaptation methods, the evident improvements of classification performance demonstrate the feasibility and effectiveness of our method.

Background Electroencephalogram (EEG) is widely used in emotion recognition due to its precision and reliability. However, the nonstationarity of EEG signals causes significant differences between individuals or sessions, making it challenging to construct a robust model. Recently, domain adaptation (DA) methods have shown excellent results in cross-subject EEG emotion recognition by aligning marginal distributions. Nevertheless, these methods do not consider emotion category labels, which can lead to label confusion during alignment. Our study aims to alleviate this problem by promoting conditional distribution alignment during domain adaptation to improve cross-subject and cross-session emotion recognition performance. Method This study introduces a multi-source domain adaptation common-branch network for EEG emotion recognition and proposes a novel sample hybridization method. This method enables the introduction of target domain data information by directionally hybridizing source and target domain samples without increasing the overall sample size, thereby enhancing the effectiveness of conditional distribution alignment in domain adaptation. Cross-subject and cross-session experiments were conducted on two publicly available datasets, SEED and SEED-IV, to validate the proposed model. Result In cross-subject emotion recognition, our method achieved an average accuracy of 90.27% on the SEED dataset, with eight out of 15 subjects attaining a recognition accuracy higher than 90%. For the SEED-IV dataset, the recognition accuracy also reached 73.21%. Additionally, in the cross-session experiment, we sequentially used two out of the three session data as source domains and the remaining session as the target domain for emotion recognition. The proposed model yielded average accuracies of 94.16 and 75.05% on the two datasets, respectively. Conclusion Our proposed method aims to alleviate the difficulties of emotion recognition from the limited generalization ability of EEG features across subjects and sessions. Though adapting the multi-source domain adaptation and the sample hybridization method, the proposed method can effectively transfer the emotion-related knowledge of known subjects and achieve accurate emotion recognition on unlabeled subjects.

For privacy protection of subjects in electroencephalogram (EEG)-based brain-computer interfaces (BCIs), using source-free domain adaptation (SFDA) for cross-subject recognition has proven to be highly effective. However, updating and storing a model trained on source subjects for each new subject can be inconvenient. This paper extends Euclidean alignment (EA) to propose adaptive Euclidean alignment (AEA), which learns a projection matrix to align the distribution of the target subject with the source subjects, thus eliminating domain drift issues and improving model classification performance of subject-independent BCIs. Combining the proposed AEA with various existing SFDA methods, such as SHOT, GSFDA, and NRC, this paper presents three new methods: AEA-SHOT, AEA-GSFDA, and AEA-NRC. In our experimental studies, these AEA-based SFDA methods were applied to four well-known deep learning models (i.e., EEGNet, Shallow ConvNet, Deep ConvNet, and MSFBCNN) on two motor imagery (MI) datasets, one event-related potential (ERP) dataset and one steady-state visual evoked potentials (SSVEP) dataset. The advanced cross-subject EEG classification performance demonstrates the efficacy of our proposed methods. For example, AEA-SHOT achieved the best average accuracy of 81.4% on the PhysioNet dataset.

Electroencephalogram (EEG) data annotation demands considerable expertise and is a time-intensive process. Moreover, inter-subject variability intensifies the challenge of domain shift, adversely impacting the generalization performance of deep learning models on unseen subjects. Current methods in EEG data analysis often struggle to handle the complex nature of brain activity without relying on EEG feature engineering. In this paper, we present a hybrid semi-supervised framework for seizure type classification, which relies on minimal domain knowledge provided by exploiting spectral and spatial patch-level representations of raw unlabeled EEG data, while leveraging a small amount of labeled data. Our method, IMPRESS, enhances EEG representation learning by combining multi-patch mutual information maximization with adversarial distribution alignment. We assessed the framework’s performance for cross-patient seizure classification using publicly accessible Temple University Seizure Corpus. IMPRESS surpasses the best-performing semi-supervised learning method by 1.92% and 0.72% using balanced accuracy and macro-F1 metrics, respectively, with 40 labeled samples per class. Remarkably, IMPRESS surpasses the fully-supervised method while requiring only 25 labeled samples per class. Additionally, we visualize the learned feature embeddings, highlighting the underlying dynamics across different seizure types, aiding in understanding the model’s behavior. This demonstrates the potential of leveraging multi-patch information from unlabeled data through a contrastive data-driven approach, alleviating the burden of annotating large amounts of EEG data.

Electroencephalogram (EEG)-based emotion recognition has gradually become a research hotspot. However, the large distribution differences of EEG signals across subjects make the current research stuck in a dilemma. To resolve this problem, in this article, we propose a novel and effective method, Multi-Source Feature Representation and Alignment Network (MS-FRAN). The effectiveness of proposed method mainly comes from three new modules: Wide Feature Extractor (WFE) for feature learning, Random Matching Operation (RMO) for model training, and Top-$\mathit{h}$ ranked domain classifier selection (TOP) for emotion classification. MS-FRAN is not only effective in aligning the distributions of each pair of source and target domains, but also capable of reducing the distributional differences among the multiple source domains. Experimental results on the public benchmark datasets SEED and DEAP have demonstrated the advantage of our method over the related competitive approaches for cross-subject EEG-based emotion recognition.

Conventional classification approaches for EEG- based emotion recognition cannot often adapt to different domains, such as cross-subject or cross-dataset scenarios, leading to poor performance. To handle this challenge, we introduce a novel fusion method using a combination of multiple domain adaptation techniques to improve the emotional states in EEG datasets via classification accuracy. For this aim, Our proposed approach exploits domain adaptation approaches such as Transfer Component Analysis (TCA), Correlation Alignment (CORAL), Transfer Joint Matching (TJM), Geodesic Flow Kernel (GFK), and Joint Distribution Adaptation (JDA), to enhance the overall classification performance. Later, a new fusion approach called Multiple Domain Adaptation based on a Neuro-Fuzzy Inference System (MDA-NF) is applied to combine the classifiers using proper fuzzy membership functions and deliver maximum separation between classes. The main contribution is by applying the fusion approach using MDA- NF technique, adaptability is sufficiently enhanced. Another advantage is to employ multiple adaptation techniques that improve separation between classes. In experimental test results conducted with cross-subject and cross-dataset scenarios, the MDA-NF approach demonstrates superior performance in terms of accuracy for both the valence and arousal aspects, as observed in two public DEAP and DREAMER datasets.

Individual differences often appear in electroencephalography (EEG) data collected from different subjects due to its weak, nonstationary and low signal-to-noise ratio properties. This causes many machine learning methods to have poor generalization performance because the independent identically distributed assumption is no longer valid in cross-subject EEG data. To this end, transfer learning has been introduced to alleviate the data distribution difference between subjects. However, most of the existing methods have focused only on domain adaptation and failed to achieve effective collaboration with label estimation. In this paper, an EEG feature transfer method combined with semi-supervised regression and bipartite graph label propagation (TSRBG) is proposed to realize the unified joint optimization of EEG feature distribution alignment and semi-supervised joint label estimation. Through the cross-subject emotion recognition experiments on the SEED-IV data set, the results show that (1) TSRBG has significantly better recognition performance in comparison with the state-of-the-art models; (2) the EEG feature distribution differences between subjects are significantly minimized in the learned shared subspace, indicating the effectiveness of domain adaptation; (3) the key EEG frequency bands and channels for cross-subject EEG emotion recognition are achieved by investigating the learned subspace, which provides more insights into the study of EEG emotion activation patterns.

Due to substantial individual variability in EEG signals, cross-subject EEG emotion recognition often suffers from poor generalizability. Although domain adaptation is widely used, single-source domain adaptation neglects the heterogeneity among distinct source subjects, while multi-source domain adaptation suffers from domain conflicts and negative transfer effects when managing multiple heterogeneous source domains simultaneously. To address these issues, we propose an augmented multi-source domain adaptation model based on TreePurgeCluster (TPC-AMDA), which incorporates TreePurgeCluster for source domain clustering and combines augmented multisource domain adversarial learning to mitigate domain conflicts, reduce computational overhead, and improve model stability. Specifically, we first employ EEG-Mixup data augmentation to generate more diverse feature samples. Next, we propose the TreePurgeCluster method to cluster different source subjects into multiple source domains, preserving the core characteristics of each domain while reducing excessive inter-domain differences. Finally, we perform multi-source domain adaptation, aligning each source domain with the target domain to minimize domain divergence and enhance the generalization ability of the model. Experiments on three benchmark datasets validate the superior performance of our approach achieving an accuracy of 93.70 % on SEED, 80.80% on SEED-IV, and 65.01% and 70.36% on the valence and arousal dimensions of DEAP, respectively.

Significant inter-individual variability limits the generalization of EEG-based emotion recognition under cross-domain settings. We address two core challenges in multi-source adaptation: (1) dynamically modeling distributional heterogeneity across sources and quantifying their relevance to a target to reduce negative transfer; and (2) achieving fine-grained semantic consistency to strengthen class discrimination. We propose a distribution-aware multi-source domain adaptation network (DAMSDAN). DAMSDAN integrates prototype-based constraints with adversarial learning to drive the encoder toward discriminative, domain-invariant emotion representations. A domain-aware source weighting strategy based on maximum mean discrepancy (MMD) dynamically estimates inter-domain shifts and reweights source contributions. In addition, a prototype-guided conditional alignment module with dual pseudo-label interaction enhances pseudo-label reliability and enables category-level, fine-grained alignment, mitigating noise propagation and semantic drift. Experiments on SEED and SEED-IV show average accuracies of 94.86\% and 79.78\% for cross-subject, and 95.12\% and 83.15\% for cross-session protocols. On the large-scale FACED dataset, DAMSDAN achieves 82.88\% (cross-subject). Extensive ablations and interpretability analyses corroborate the effectiveness of the proposed framework for cross-domain EEG-based emotion recognition.

No abstract available

Motor Imagery (MI) paradigm is critical in neural rehabilitation and gaming. Advances in brain-computer interface (BCI) technology have facilitated the detection of MI from electroencephalogram (EEG). Previous studies have proposed various EEG-based classification algorithms to identify the MI, however, the performance of prior models was limited due to the cross-subject heterogeneity in EEG data and the shortage of EEG data for training. Therefore, inspired by generative adversarial network (GAN), this study aims to propose an improved domain adaption network based on Wasserstein distance, which utilizes existing labeled data from multiple subjects (source domain) to improve the performance of MI classification on a single subject (target domain). Specifically, our proposed framework consists of three components, including a feature extractor, a domain discriminator, and a classifier. The feature extractor employs an attention mechanism and a variance layer to improve the discrimination of features extracted from different MI classes. Next, the domain discriminator adopts the Wasserstein matrix to measure the distance between source domain and target domain, and aligns the data distributions of source and target domain via adversarial learning strategy. Finally, the classifier uses the knowledge acquired from the source domain to predict the labels in the target domain. The proposed EEG-based MI classification framework was evaluated by two open-source datasets, the BCI Competition IV Datasets 2a and 2b. Our results demonstrated that the proposed framework could enhance the performance of EEG-based MI detection, achieving better classification results compared with several state-of-the-art algorithms. In conclusion, this study is promising in helping the neural rehabilitation of different neuropsychiatric diseases.

Robust decoding performance is essential for the practical deployment of brain-computer interface (BCI) systems. Existing EEG decoding models often rely on large amounts of annotated data collected through specific experimental setups, which fail to address the heterogeneity of data distributions across different domains. This limitation hinders BCI systems from effectively managing the complexity and variability of real-world data. To overcome these challenges, we propose Synchronized Self-Training Domain Adaptation (SSTDA) for cross-domain motor imagery classification. Specifically, SSTDA leverages labeled signals from a source domain and applies self-training to unlabeled signals from a target domain, enabling the simultaneous training of a more robust classifier. The raw EEG signals are mapped into a latent space by a feature extractor for discriminative representation learning. A domain-shared latent space is then learned by optimizing the feature extractor with both source and target samples, using an easy-tohard self-training process. We validate the method with extensive experiments on two public motor imagery datasets: Dataset IIa of BCI Competition IV and the High Gamma dataset. In the inter-subject task, our method achieves classification accuracies of 64.43% and 80.40%, respectively. It also outperforms existing methods in the inter-session task. Moreover, we develope a new six-class motor imagery dataset and achieve test accuracies of 77.09% and 80.18% across different datasets. All experimental results demonstrate that our SSTDA outperforms existing algorithms in inter-session, inter-subject, and inter-dataset validation protocols, highlighting its capability to learn discriminative, domain-invariant representations that enhance EEG decoding performance.

The variability and dynamic nature of electroencephalography (EEG) signals pose challenges in the analysis and classification of cross-domain data for EEG emotion recognition. To tackle the heterogeneity and distributional inconsistencies inherent in multi-source EEG data, we introduce a Multi-Source Domain Adversarial Training Network (MD-ATN). This approach utilizes a multi-branch feature extraction architecture to capture shared features across multiple source domains, while incorporating a Gradient Reversal Layer (GRL) for adversarial training between feature extractors and domain classifiers to generate domain-invariant generic features. Moreover, our model implements a joint optimization strategy that combines Maximum Mean Discrepancy (MMD) and Distribution Discrepancy Loss (DDL) to enhance domain adaptation. Experimental results demonstrate that MD-ATN outperforms existing methods across SEED datasets. Specifically, the MD-ATN model achieves an average accuracy of 87.1 % in the Cross-Subject task and 89.5% in the Cross-Session task.

Brain computer interface (BCI) provides a promising and intelligent rehabilitation method for motor function, and it is crucial to acquire the patient’s movement intentions accurately through decoding motor imagery EEG (MI-EEG) . Because of the inter-individual heterogeneity, the decoding model should demonstrate dynamic adaptation abilities.Domain adaptation (DA) is effective to enhance model generalization by reducing the inherent distribution difference among subjects. However, the existing DA methods usually mix the multiple source domains into a new domain, the resulting multi-source domain conflict may cause negative transfer. In this paper, we propose a multi-source dynamic conditional domain adaptation network (MSDCDA). First, a multi-channel attention block is employed in the feature extractor to focus on the channels relevant to the corresponding MI task. Subsequently, the shallow spatial-temporal features are extracted using a spatial-temporal convolution block. And a dynamic residual block is applied in the feature extractor to dynamically adapt specific features of each subject to alleviate conflicts among multiple source domains since each domain is viewed as a distribution of electroencephalogram (EEG) signals. Furthermore, we employ the Margin Disparity Discrepancy (MDD) as the metric to achieve conditional distribution domain adaptation between the source and target domains through adversarial learning with an auxiliary classifier. MSDCDA achieved accuracies of 78.55%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document} and 85.08%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document} on Datasets IIa and IIb of BCI Competition IV, respectively. Our experimental results demonstrate that MSDCDA can effectively address multi-source domain conflicts and significantly enhance the decoding performance of target subjects. This study positively facilitates the application of BCI based on motor function rehabilitation.

Emotion recognition is crucial for advancing mental health, healthcare, and technologies like brain-computer interfaces (BCIs). However, EEG-based emotion recognition models face challenges in cross-domain applications due to the high cost of labeled data and variations in EEG signals from individual differences and recording conditions. Unsupervised domain adaptation methods typically require access to source domain data, which may not always be feasible in real-world scenarios due to privacy and computational constraints. Source-free unsupervised domain adaptation (SF-UDA) has recently emerged as a solution, enabling target domain adaptation without source data, but its application in emotion recognition remains unexplored. We propose a novel SF-UDA approach for EEG-based emotion classification across domains, introducing a multi-stage framework that enhances model adaptability without requiring source data. Our approach incorporates Dual-Loss Adaptive Regularization (DLAR) to minimize prediction discrepancies on confident samples and align predictions with expected pseudo-labels. Additionally, we introduce Localized Consistency Learning (LCL), which enforces local consistency by promoting similar predictions from reliable neighbors. These techniques together address domain shift and reduce the impact of noisy pseudo-labels, a key challenge in traditional SF-UDA models. Experiments on two widely used datasets, DEAP and SEED, demonstrate the effectiveness of our method. Our approach significantly outperforms state-of-the-art methods, achieving 65.84% accuracy when trained on DEAP and tested on SEED, and 58.99% accuracy in the reverse scenario. It excels at detecting both positive and negative emotions, making it well-suited for practical emotion recognition applications.

EEG emotion recognition is crucial in both human-machine interaction and healthcare. However, recognizing emotions across different subjects remains challenging due to individual variability. While existing multi-source domain adaptation methods have been utilized for cross-subject EEG emotion decoding, they often struggle with irrelevant or weakly relevant source domains, leading to negative transfer. Additionally, variations within subdomains are often neglected in these studies. We propose a joint domain adaptation method, Adaptive Source Joint Domain Adaptation (ASJDA) to address these issues. ASJDA utilizes an unsupervised adaptive source selection strategy to select a subset of source domains by evaluating the Jensen-Shannon divergence between the source and target domains, choosing those most relevant to the target. Subsequently, it implements joint domain adaptation with these chosen sources at both the domain and category subdomain levels. Our proposed method outperforms existing state-of-the-art methods, achieving cross-subject accuracies of 96.81% in SEED, 89.69% in SEED-IV, and 69.31% in DEAP. This work significantly advances the state of the art in EEG emotion recognition by effectively addressing the challenges of cross-subject variability.

No abstract available

No abstract available

Due to the inherent non-stationarity and individual differences present in electroencephalogram (EEG) signals, developing a generalizable model that performs well on new subjects is challenging in EEG-based emotion recognition. Most existing domain adaptation (DA) methods typically mitigate these discrepancies by aligning the marginal distributions of domain feature representations. However, when there is a significant difference in the class-conditional distribution between domain features and labels, the domain-invariant features learned by aligning marginal distributions may have limited discriminative ability for unlabeled target instances or even prove counterproductive. To address this issue, we propose a Neighborhood Semantic Aware Learning-based Dynamic Graph Attention Convolution (NSAL-DGAT) approach that learns target semantic information by considering the inter-domain semantic topological structure, thereby improving classifier adaptation for target instances. Specifically, the proposed NSAL framework is designed to capitalize on the insight that after domain feature alignment, some target samples and their neighboring source samples exhibit similar semantics. By leveraging the neighborhood topological structure, we extract and incorporate semantic target features to train a more transferable classifier. Besides, we implement an entropy weighting mechanism to emphasize representative target semantic information, encouraging target instances to prioritize high-confidence individuals within the source neighborhood. We have conducted extensive experiments on the public SEED dataset and our collected the Hearing-Impaired EEG Dataset (HIED). The experimental results underscore the efficacy of our proposed NSAL-DGAT approach, showcasing state-of-the-art accuracy in subject-dependent as well as subject-independent scenarios.

Emotion recognition from electroencephalography (EEG) signals is increasingly emerging as a critical research focus in brain-computer interfaces (BCIs). However, challenges such as the scarcity of emotion labels and distribution discrepancies in EEG signals significantly hinder the practical application of EEG-based emotion recognition. To overcome these challenges, this article fully exploits the continuity of emotion-related EEG data and proposes an unsupervised domain adaptation (DA) with pseudo-label propagation (PLP), termed DA method combined with PLP (DAPLP), for cross-domain EEG emotion recognition. Specifically, we first perform global distribution alignment (GDA) between the source and target domains and utilize the source classifier to generate pseudo-labels for the target domain. From these predictions, reliable pseudo-labels are then selected to guide label propagation, and the propagation process is further optimized with correct and smooth techniques. Systematic experiments conducted on the SEED, SEED-IV, and SEED-V datasets reveal that the proposed DAPLP accomplishes competitive performance compared to advanced existing methods, reaching average accuracies of 89.44%/74.57%/69.15% in cross-subject evaluation and 96.41%/82.20%/84.70% in cross-session evaluation, respectively. Moreover, our proposed DAPLP exhibits strong practical potential and robust performance in unsupervised cross-domain emotion recognition.

In electroencephalographic-based (EEG-based) emotion recognition, high non-stationarity and individual differences in EEG signals could lead to significant discrepancies between sessions/subjects, making generalization to a new session/subject very difficult. Most existing domain adaptation (DA) and multi-source domain adaptation (MSDA) techniques aim to mitigate this discrepancy by aligning feature distributions. However, when confronted with many diverse domain distributions, learning domain-invariant features via aligning pairwise feature distributions between domains can be hard or even counterproductive. To address this issue, this article proposes an attention alignment approach to learning abundant domain-invariant features. The motivation is simple: despite individual differences causing significant differences in feature distributions in EEG-based emotion recognition, shared affective cognitive attributes (attention) of spectral and spatial domains can be observed within the same emotion categories. The proposed spectral-spatial attention alignment multi-source domain adaptation (S2A2-MSDA) constructs domain attention to represent affective cognition attributes in spatial and spectral domains and utilizes domain consistent loss to align them between domains. Furthermore, to facilitate discriminative feature learning on the target classes, S2A2-MSDA learns the conditional semantic information of the target domain using a pseudo-labeling method. This algorithm has been validated on the SEED and SEED-IV datasets in cross-session and cross-subject scenarios, respectively. Experimental results demonstrate that S2A2-MSDA outperforms existing representative DA and MSDA methods, achieving state-of-the-art performance.

No abstract available

Affective brain-computer interface is an important part of realizing emotional human-computer interaction. However, existing objective individual differences among subjects significantly hinder the application of electroencephalography (EEG) emotion recognition. Existing methods still lack the complete extraction of subject-invariant representations for EEG and the ability to fuse valuable information from multiple subjects to facilitate the emotion recognition of the target subject. To address the above challenges, we propose a Multi-source Selective Graph Domain Adaptation Network (MSGDAN), which can better utilize data from different source subjects and perform more robust emotion recognition on the target subject. The proposed network extracts and selects the individual information specific to each subject, where public information refers to subject-invariant components from multi-source subjects. Moreover, the graph domain adaptation network captures both functional connectivity and regional states of the brain via a dynamic graph network and then integrates graph domain adaptation to ensure the invariance of both functional connectivity and regional states. To evaluate our method, we conduct cross-subject emotion recognition experiments on the SEED, SEED-IV, and DEAP datasets. The results demonstrate that the MSGDAN has superior classification performance.

Electroencephalogram (EEG) is widely utilized in emotion recognition owing to its unique advantages. To achieve more optimal cross-subject emotion recognition, a cross subject emotion recognition method based on interconnection dynamic domain adaptation (IDDA) is proposed. In IDDA, dynamic graph convolution (DGC) is employed to dynamically learn the intrinsic relationships between different EEG channels and to extract domain invariant features. And dynamic domain adaptation (DDA) is employed to align the source domain and target domain, at the same time the emotional sub-domains is aligned, achieving more optimal cross subject emotion recognition. To select suitable subjects as the source domain, a multi-source selection algorithm is incorporated before dynamic adaptive computation reducing migration noise and achieving interconnection between DGC and DDA. IDDA enhances the emotion discrimination ability of domain invariant features, thereby improving the accuracy of cross-subject EEG emotion recognition. This method achieves classification results of 85.75% and 72.36% in cross subject experiments on SEED and SEED-IV.

Emotion recognition based on electroencephalography (EEG) has significant advantages in terms of reliability and accuracy. However, individual differences in EEG limit the ability of sentiment classifiers to generalize across subjects. Furthermore, due to the nonstationarity of EEG, subject signals can vary with time, an important challenge for temporal emotion recognition. Several emotion recognition methods have been developed that consider the alignment of conditional distributions, but do not balance the weights of conditional and marginal distributions. In this article, we propose a novel approach to generalize emotion recognition models across individuals and time, i.e., global and local associative domain adaptation (GLADA). The proposed method consists of three parts: 1) deep neural networks are used to extract deep features from emotional EEG data; 2) considering that marginal and conditional distributions between domains can contribute to adaptation differently, a method that combines coarse-grained adversarial adaptation and fine-grained adversarial adaptation is used to narrow the domain distance of the joint distribution in the EEG data between subjects (i.e., reduce intersubject variability), and the weights of the marginal and conditional distributions are automatically balanced using dynamic balancing factors; and 3) domain adaptation is used to accelerate model convergence. Using GLADA, subject-independent EEG emotion recognition is improved by reducing the influence of the subject’s personal information on EEG emotion. Experimental results demonstrate that the GLADA model effectively addresses the domain transfer problem, resulting in improved performance across multiple EEG emotion recognition tasks.

Research on emotion recognition based on EEG signals has made significant progress. Most of the existing studies have focused on supervised learning methods, but real-life data cannot meet the requirement of high quality with labels. In addition, EEG signals have individual variability and instability, which requires transfer learning to enhance the model generalization. In this paper, we propose a multi-view self-supervised domain adaptation model that combines self-supervised learning techniques with domain-adaptive transfer learning algorithms, which can solve the last two problems mentioned above. Specifically, we add a multi-class domain discriminator to construct the adversarial relationship between the sub-networks so that distribution discrepancy of different subjects can be reduced effectively. We conduct both subject-dependent and subject-independent experiments on the SEED and SEED-IV datasets to thoroughly evaluate the performance of our model. The results show that our model achieves outstanding emotion recognition performance even with limited labeled data. In the subject-dependent experiments on both datasets, our model achieves accuracy rates of 85.91% and 87.19% respectively, surpassing the original self-supervised masked autoencoder model by about 3%. In subject-independent experiments, our model demonstrates strong data distribution adaptation capabilities, achieving an accuracy of 69.72% and 62.87%, respectively on the SEED and SEED-IV datasets using only 90 samples for subject-independent experiments. This effectively mitigates the accuracy degradation caused by differences in data distribution across subjects. Furthermore, our model is capable of extracting meaningful features from corrupted EEG data, highlighting its robustness and effectiveness.

The variability of EEG signals produced by different trial conditions and different devices presents significant challenges in developing practical EEG-based emotion recognition systems. Much of the research on developing a generalizable EEG emotion recognition approach focuses on cross-subject and cross-session contexts. Although current cross-subject methods yield satisfactory outcomes on certain EEG emotion datasets, their effectiveness diminishes in cross-dataset scenarios. Additionally, the performance of existing cross-dataset methods remains inferior compared to methods that are trained and evaluated within the same dataset. To address these challenges, inspired by the effective application of deep metric learning (DML) in zero-shot and few-shot learning tasks, this paper introduces a cross-dataset emotion recognition method. The proposed approach integrates DML, domain-specific batch normalization (DSBN), shared batch normalization statistics, and adversarial learning. Specifically, our method extracts cross-domain features from the input signals using DSBN and shared batch normalization statistics. Then, the proposed DML loss minimizes intra-class variations of EEG features across different subjects and domains, while maximizing differences between different classes. Moreover, it captures the semantic order of emotions in the learned embedding. To further improve the generalization of the feature encoder, we employ adversarial learning with domain and subject discriminators. We evaluate our method on six cross-dataset scenarios. The results show that it consistently outperforms peer methods across the scenarios. For example, our method achieves an accuracy of 63.49% on SEED $\to $ SEED-IV, improving the state-of-the-art result by 2.25%.

This paper introduces an Emotion Domain Adversarial Neural Network (EDANN) model for Electroencephalogram (EEG) emotion recognition, designed to accomplish EEG emotion classification across various periods and subjects. The model is composed of three essential components: an encoder, a label classifier, and a domain discriminator. Utilizing adversarial training, EDANN can extract features that are discriminative among different categories and invariant across domains. The study employed two public datasets: the DEAP dataset, which includes EEG signals from 32 participants for emotion analysis; and the SEED dataset, comprising emotional responses from 15 Chinese participants to 15 Chinese film clips. Both datasets utilized specific emotion models for labeling emotions, albeit with varying levels of precision. Extensive experiments on both Session-to-Session and Subject-to-Subject transfer tasks have demonstrated the proposed model’s superior performance in terms of accuracy, precision, recall, and F1 score for emotion recognition. The findings of this study not only illustrate the efficacy of EDANN in EEG-based emotion recognition but also underscore the importance of considering significant inter-individual differences when designing and evaluating machine learning models. These methodologies enable researchers to utilize limited data resources more efficiently, thus propelling the advancement of emotion recognition technology.

Transfer learning (TL) has been employed in electroencephalogram (EEG)-based brain–computer interfaces (BCIs) to enhance performance for cross-session and cross-subject EEG classification. However, domain shifts coupled with a low signal-to-noise ratio between EEG recordings have been demonstrated to contribute to significant variations in EEG neural dynamics from session to session and subject to subject. Critical factors—such as mental fatigue, concentration, and physiological and non-physiological artifacts—can constitute the immense domain shifts seen between EEG recordings, leading to massive inter-subject variations. Consequently, such variations increase the distribution shifts across the source and target domains, in turn weakening the discriminative knowledge of classes and resulting in poor cross-subject transfer performance. In this paper, domain adaptation algorithms, including two machine learning (ML) algorithms, are contrasted based on the single-source-to-single-target (STS) and multi-source-to-single-target (MTS) transfer paradigms, mainly to mitigate the challenge of immense inter-subject variations in EEG neural dynamics that lead to poor classification performance. Afterward, we evaluate the effect of the STS and MTS transfer paradigms on cross-subject transfer performance utilizing three EEG datasets. In this case, to evaluate the effect of STS and MTS transfer schemes on classification performance, domain adaptation algorithms (DAA)—including ML algorithms implemented through a traditional BCI—are compared, namely, manifold embedded knowledge transfer (MEKT), multi-source manifold feature transfer learning (MMFT), k-nearest neighbor (K-NN), and Naïve Bayes (NB). The experimental results illustrated that compared to traditional ML methods, DAA can significantly reduce immense variations in EEG characteristics, in turn resulting in superior cross-subject transfer performance. Notably, superior classification accuracies (CAs) were noted when MMFT was applied, with mean CAs of 89% and 83% recorded, while MEKT recorded mean CAs of 87% and 76% under the STS and MTS transfer paradigms, respectively.

Electroencephalogram (EEG)-based emotion recognition is a vital component in brain-computer interface applications. However, it faces two significant challenges: 1) extracting domain-invariant features while effectively preserving emotion-related information, and 2) aligning the joint probability distributions of data across different individuals. To address these challenges, we propose a progressive multi-domain adaptation network with reinforced self-constructed graphs. Specifically, we introduce EEG-CutMix to construct unlabeled mixed-domain data, facilitating the transition between source and target domains. Additionally, a reinforced self-constructed graphs module is employed to extract domain-invariant features. Finally, a progressive multi-domain adaptation framework is constructed to smoothly align the data distributions across individuals. Experiments on cross-subject datasets demonstrate that our model achieves state-of-the-art performance on the SEED and SEED-IV datasets, with accuracies of 97.03% <inline-formula> <tex-math notation="LaTeX">$\pm ~1.65$ </tex-math></inline-formula>% and 88.18% <inline-formula> <tex-math notation="LaTeX">$\pm ~4.55$ </tex-math></inline-formula>%, respectively. Furthermore, tests on a self-recorded dataset, comprising ten healthy subjects and twelve patients with disorders of consciousness (DOC), show that our model achieves a mean accuracy of 86.65% <inline-formula> <tex-math notation="LaTeX">$\pm ~2.28$ </tex-math></inline-formula>% in healthy subjects. Notably, it successfully applies to DOC patients, with four subjects achieving emotion recognition accuracy exceeding 70%. These results validate the effectiveness of our model in EEG emotion recognition and highlight its potential for assessing consciousness levels in DOC patients. The source code for the proposed model is available at GitHub-seizeall/mycode.

Recognizing emotions using EEG signals is difficult because EEG data is not stationary, has a low signal-to-noise ratio, and varies a lot between subjects. We present a new hybrid framework called CDA-GAF (Cross-Domain Adaptive Graph Attention Fusion) in this work. It combines the strengths of Graph Attention Networks (GATs), Temporal Transformers, and Domain Adaptation to make emotion classification models more robust and generalizable. To make brain connectivity graphs for each frequency band, our method first gets functional connectivity features from EEG channels. A GAT module processes these to find spatial dependencies in EEG activity. Then, a Temporal Transformer module is used to model long-range dependencies between EEG sequences. To address cross-subject variations, we implement a domain adaptation layer utilizing CORAL loss or Domain-Adversarial Training (DANN), which aligns feature distributions between source and target subjects. We also use extra emotion supervision signals, like HRV or micro-expressions, to improve the quality of the labels by anchoring the emotional state in multiple ways. We test our model on standard datasets like DEAP, SEED, and WESAD. It does much better than baseline models at recognizing emotions in both within-subject and cross-subject settings. Our findings underscore the efficacy of integrating graph-based spatial encoding, temporal attention mechanisms, and domain adaptation for emotion recognition from EEG data.

No abstract available

Electroencephalography (EEG) has emerged as a widely utilized signal in motor imagery (MI) brain-computer interfaces(BCI) due to its convenience and safety. Recently, deep learning methods have rapidly developed in the field of brain computer interfaces. However, traditional EEG classification methods often face challenges related to limited generalization capability across subjects. To address this issue, this paper proposes a multi-source discriminant dynamic domain adaptation model(MSD-DDA) aimed at fully leveraging domain adaptation to enhance the accuracy of motor imagery classification. The model adeptly handles global and local disparities in motor imagery classification by dynamically minimizing differences between global domain and local subdomain. Furthermore, to ensure discriminability and diversity in the target domain, we introduce batch kernel norm maximization of the difference, thereby enhancing the model's discriminability in the target domain while preserving prediction diversity. To tackle variations in similarity between different source domains and the target domain, we devise a weighted joint prediction mechanism. This mechanism automatically adjusts the contribution weight of each source domain based on its similarity to the target domain, facilitating more precise discriminant prediction and improved adaptability to scenarios with multiple source domains. To evaluate our approach, we conducted a large number of experiments on datasets 1 and 2a of the Fourth BCI Competition and on the openBMI dataset, with average classification accuracy of 92.43%, 79.24% and 71.96%, respectively.Finally, we compare the proposed method with several classical and recent algorithms, and prove that its performance is better than the existing methods.

EEG-based emotion recognition (EEG-ER) through deep learning models has gained more attention in recent years, with more researchers focusing on architecture, feature extraction, and generalisability. This paper presents a novel end-to-end deep learning framework for EEG-ER, combining temporal feature extraction, self-attention mechanisms, and adversarial domain adaptation. The architecture entails a multi-stage 1D CNN for spatiotemporal features from raw EEG signals, followed by a transformer-based attention module for long-range dependencies, and a domain-adversarial neural network (DANN) module with gradient reversal to enable a powerful subject-independent generalisation by learning domain-invariant features. Experiments on benchmark datasets (DEAP, SEED, DREAMER) demonstrate that our approach achieves a state-of-the-art performance, with a significant improvement in cross-subject recognition accuracy compared to non-adaptive frameworks. The architecture tackles key challenges in EEG emotion recognition, including generalisability, inter-subject variability, and temporal dynamics modelling. The results highlight the effectiveness of combining convolutional feature learning with adversarial domain adaptation for robust EEG-ER.

No abstract available

No abstract available

In this paper, we focus on the challenge of individual variability in affective brain-computer interfaces (aBCI), which employs electroencephalogram (EEG) signals to monitor and recognize human emotional states, thereby facilitating the advancement of emotion-aware technologies. The variability in EEG data across individuals poses a significant barrier to the development of effective and widely applicable aBCI models. To tackle this issue, we propose a novel transfer learning framework called Semi-supervised Domain Adaptation with Dynamic Distribution Alignment (SDA-DDA). This approach aligns the marginal and conditional probability distribution of source and target domains using maximum mean discrepancy (MMD) and conditional maximum mean discrepancy (CMMD). We introduce a dynamic distribution alignment mechanism to adjust differences throughout training and enhance adaptation. Additionally, a pseudo-label confidence filtering module is integrated into the semi-supervised process to refine pseudo-label generation and improve the estimation of conditional distributions. Extensive experiments on EEG benchmark databases (SEED, SEED-IV and DEAP) validate the robustness and effectiveness of SDA-DDA. The results demonstrate its superiority over existing methods in emotion recognition across various scenarios, including cross-subject and cross-session conditions. This advancement enhances the generalization and accuracy of emotion recognition, potentially fostering the development of personalized aBCI applications. The source code is accessible at https://github.com/XuanSuTrum/SDA-DDA.

Objective. Accurate sleep-stage classification is crucial for advancing both sleep research and healthcare applications. Traditional deep learning (DL) and domain adaptation (DA) methods often struggle due to the limited availability of labeled data in the target domain and their inability to capture the subtle distinctions between sleep-stage classes, which hampers classification accuracy. Approach. To address these limitations, we introduce a novel framework, adversarial domain adaptation with active deep learning (ADAADL). This framework combines adversarial learning with active learning (AL) strategies to improve feature alignment and effectively leverage unlabeled data. ADAADL employs two separate sleep-stage classifiers as discriminators, allowing for a more refined consideration of class boundaries during the feature alignment process. Moreover, it incorporates entropy measures alongside cross-entropy loss during training to make better use of the information from unlabeled data. The AL component (ADL) further enhances performance by iteratively selecting and labeling the most informative data points, thereby reducing annotation efforts and improving generalization to unseen data. Main results. Experimental evaluations on three benchmark EEG datasets demonstrate that ADAADL produces robust, transferable features, significantly outperforming existing DA methods in classification accuracy. This research advances sleep-stage classification techniques, offering improved accuracy for real-world applications and contributing to a deeper understanding of sleep dynamics. Significance. The proposed ADAADL framework advances the state of the art in sleep-stage classification by effectively leveraging unlabeled data and reducing labeling costs. It offers a scalable and accurate solution for real-world sleep monitoring applications and contributes to a deeper understanding of sleep dynamics through improved modeling of sleep stages.

Cross-session pose a challenge to the application of EEG-based emotion recognition (ER), which is due to the non-stationary nature of EEG that causes the EEG to reveal distribution discrepancies over time, thereby leading to degradation of performance. The traditional way is by collecting labelled data over multiple sessions and then retraining a new model, but this is time-consuming and labor-intensive. In this paper, we propose the Maximizing Domain Discrepancy for EEG-based ER (MDD-ER) to improve cross-session performance. MDD-ER applies distinct domain adaptation strategies to alleviate feature distribution discrepancies between source session and target session at different levels: for shallow features, we use maximum mean discrepancy (MMD) to align the source and target domains based on statistical criterion, and for deep features, we adversarially train two classifiers, which effectively improves the alignment precision of the source and target domains due to the consideration of the class of features. Consequently, the proposed MDD-ER method can improve model generalization across sessions. We conduct comprehensive experiments on SEED dataset, and the experimental results demonstrate the effectiveness of the proposed MDD-ER method in cross-session ER.

No abstract available

Without human participation in driving operations, the adoption of autonomous driving (AD) technology greatly enhances driving safety by reducing human errors. Even though AD can handle common scenarios properly, some exceptions still call for the human takeover with AD failing to engage due to the incomprehensible or intensely conflict situations that rarely occur. To help AD understand and recognize the disengagement scenarios effectively, this paper incorporates the human electroencephalogram (EEG) cognitive data into modeling and proposes a transfer learning framework to let AD absorb the integrative knowledge from the manual driving (MD). Several disengagement scenarios are designed using a driving simulator and EEG data are collected from both "drivers" in MD and "supervisors" in AD. A conditional maximum mean discrepancy (CMMD) function is introduced to identify the common brain activity characteristics, allowing the recognition model to be transferred from the cognitively demanding domain of MD to the less demanding domain of AD. The results indicate that the proposed model can achieve an 80 % recognition rate for typical disengagement scenarios, such as static obstacles, intersection conflict and vehicle cut-in, using only 30 % of AD training labels. The transferable common feature space from EEG data improves the recognition accuracy by 21.2 % compared with the model only using AD domain data. By accurately recognizing the type of disengagement scenarios, the AD system can activate appropriate safety mechanisms or provide more explicit takeover prompts, which could effectively reduce the risk of accidents due to delayed or incorrect takeovers.

Emotion recognition based on electroencephalogram (EEG) faces substantial challenges. The variability of neural signals among different subjects and the scarcity of labeled data pose obstacles to the generalization ability of traditional domain adaptation (DA) methods. Existing approaches, especially those relying on the maximum mean discrepancy (MMD) technique, are often highly sensitive to domain mean shifts induced by noise. To overcome these limitations, a novel framework named Domain Adaptive Deep Possibilistic clustering (DADPc) is proposed. This framework integrates deep domain-invariant feature learning with possibilistic clustering, reformulating Maximum Mean Discrepancy (MMD) as a one-centroid clustering task under a fuzzy entropy-regularized framework. Moreover, the DADPc incorporates adaptive weighted loss and memory bank strategies to enhance the reliability of pseudo-labels and cross-domain alignment. The proposed framework effectively mitigates noise-induced domain shifts while maintaining feature discriminability, offering a robust solution for EEG-based emotion recognition in practical applications. Extensive experiments conducted on three benchmark datasets (SEED, SEED-IV, and DEAP) demonstrate the superior performance of DADPc in emotion recognition tasks. The results show significant improvements in recognition accuracy and generalization capability across different experimental protocols, including cross-subject and cross-session scenarios. This research contributes to the field by providing a comprehensive approach that combines deep learning with possibilistic clustering, advancing the state-of-the-art in cross-domain EEG analysis.

Parkinson’s disease has brought great harm to human life and health. The detection of Parkinson’s disease based on electroencephalogram (EEG) provides a new way to prevent and treat Parkinson’s disease. However, due to the limited EEG data samples, there are large differences among different subjects, especially among different datasets. In this study, a new method called Improved Convex Hull and Maximum Mean Discrepancy (ICMMD)for cross-dataset classification of Parkinson’s disease is proposed by combining convex hull and transfer learning. The paper innovatively implements cross-data transfer learning in the field of brain–computer interfaces for Parkinson’s disease, using Euclidean distance for data alignment and EEG channel selection, and combines the convex envelope with MMD distance to form an effective source domain selection method. Lowpd, San and UNM datasets are used to verify the effectiveness of the proposed method through experiments on different brain regions and frequency bands in Parkinson’s. The results show that this method has good classification performance in different regions of the brain and frequency bands. The research in this paper provides a new idea and method for disease detection of Parkinson’s disease across datasets.

Dysarthria impairs motor control of speech, often resulting in reduced intelligibility and frequent misarticulations. Although interest in brain–computer interface technologies is growing, electroencephalogram (EEG)-based communication support for individuals with dysarthria remains limited. To address this gap, we recorded EEG data from one participant with dysarthria during a Korean automatic speech task and labeled each trial as correct or misarticulated. Spectral analysis revealed that misarticulated trials exhibited elevated frontal–central delta and alpha power, along with reduced temporal gamma activity. Building on these observations, we developed a soft multitask learning framework designed to suppress these nonspecific spectral responses and incorporated a maximum mean discrepancy–based alignment module to enhance class discrimination while minimizing domain-related variability. The proposed model achieved F1-scores of 52.7 % for correct and 41.4 % for misarticulated trials—an improvement of 2 % and 11 % over the baseline—demonstrating more stable intention decoding even under articulation errors. These results highlight the potential of EEG-based assistive systems for communication in language impaired individuals.

Electroencephalography (EEG) is crucial for diagnosing neurological disorders, but model generalization is often hampered by domain shifts arising from variations in hardware, patient populations, recording protocols, and annotation standards. We propose a domain adaptation model that uses EEG-specific properties, integrating the Mean Teacher Model (MTM), Maximum Mean Discrepancy (MMD), Topology Loss, Frequency Regularization, and Attractor Loss to align heterogeneous datasets.Experiments using CHB-MIT as the source and TUH Seizure / MDD datasets as targets demonstrate significant improvements in classification performance (F1 score) and domain alignment (Jaccard coefficient). The model successfully preserves critical neurophysiological markers, capturing 20Hz epilepsy-related spikes and parietal alpha suppression associated with depression. It also highlights spatial contributions from key channels like P8, constructing a latent space that effectively retains epileptic and depressive EEG features.

Detecting single-trial P300 from EEG is difficult when only a few labeled trials are available. When attempting to boost a small target set with a large source dataset through transfer learning, cross-dataset shift arises. To address this challenge, we study transfer between two public visual-oddball ERP datasets using five shared electrodes (Fz, Pz, P3, P4, Oz) under a strict small-sample regime (target: 10 trials/subject; source: 80 trials/subject). We introduce Adaptive Split Maximum Mean Discrepancy Training (AS-MMD), which combines (i) a target-weighted loss with warm-up tied to the square root of the source/target size ratio, (ii) Split Batch Normalization (Split-BN) with shared affine parameters and per-domain running statistics, and (iii) a parameter-free logit-level Radial Basis Function kernel Maximum Mean Discrepancy (RBF-MMD) term using the median-bandwidth heuristic. Implemented on an EEG Conformer, AS-MMD is backbone-agnostic and leaves the inference-time model unchanged. Across both transfer directions, it outperforms target-only and pooled training (Active Visual Oddball: accuracy/AUC 0.66/0.74; ERP CORE P3: 0.61/0.65), with gains over pooling significant under corrected paired t-tests. Ablations attribute improvements to all three components.

No abstract available

No abstract available

For solving the problem of the inevitable decline in the accuracy of cross-subject emotion recognition via Electroencephalograph (EEG) signal transfer learning due to the negative transfer of data in the source domain, this paper offers a new method to dynamically select the data suitable for transfer learning and eliminate the data that may lead to negative transfer. The method which is called cross-subject source domain selection (CSDS) consists of the next three parts. 1) First, a Frank-copula model is established according to Copula function theory to study the correlation between the source domain and the target domain, which is described by the Kendall correlation coefficient. 2) The calculation method for the Maximum Mean Discrepancy is improved to determine the distance between classes in a single source. After normalization, the Kendall correlation coefficient is superimposed, and the threshold is set to identify the source-domain data most suitable for transfer learning. 3) In the process of transfer learning, on the basis of Manifold Embedded Distribution Alignment, the Local Tangent Space Alignment method is used to provide a low-dimensional linear estimation of the local geometry of nonlinear manifolds, which maintains the local characteristics of the sample data after dimensionality reduction. Experimental results show that compared with the traditional methods, the CSDS increases the accuracy of emotion classification by approximately 2.8% and reduces the runtime by approximately 65%.

Electroencephalogram (EEG) emotion recognition plays an important role in human–computer interaction. An increasing number of algorithms for emotion recognition have been proposed recently. However, it is still challenging to make efficient use of emotional activity knowledge. In this paper, based on prior knowledge that emotion varies slowly across time, we propose a temporal-difference minimizing neural network (TDMNN) for EEG emotion recognition. We use maximum mean discrepancy (MMD) technology to evaluate the difference in EEG features across time and minimize the difference by a multibranch convolutional recurrent network. State-of-the-art performances are achieved using the proposed method on the SEED, SEED-IV, DEAP and DREAMER datasets, demonstrating the effectiveness of including prior knowledge in EEG emotion recognition.

No abstract available

The safety monitoring system of intelligent transportation provides driving fatigue warning and risk control. Electroencephalogram (EEG) signals can directly reflect the neuronal activity of the brain. The detection and early warning of driving fatigue using EEG signals has important practical significance. However, because of the non-stationarity and timeliness of EEG signals, the single feature detection method is significantly impacted by data distribution differences. In this paper, in the framework of multi-input multi-output (MIMO) Takagi-Sugeno-Kang (TSK) fuzzy system, transferable TSK fuzzy classifier with multi-views (T-TSK-MV) is developed for EEG-based driving fatigue recognition in intelligent transportation. First, in view-specific consequent parameter learning, the view-specific consequent regularizer is designed based on technologies of ridge regression, maximum mean discrepancy (MMD), and manifold regularization, which becomes the bridge to transfer the discriminative information from the related domain to the target domain. In addition, the $\ell _{2,1} $ -norm sparse constraint on consequent parameters is used to simplify fuzzy rules. Then multi-view learning is integrated into the consequent parameter learning, in which T-TSK-MV explores the view-shared consequent regularizer and adaptively assigns weights to each view. The $\ell _{2,1} $ -norm sparse constraint on view-shared consequent regularizer can effectively exploit the local structure of multi-view data. Finally, the fuzzy classifier is constructed on view-specific regularizers and view weights. The experiment on real-word datasets shows that the proposed fuzzy classifier can significantly improve the driving fatigue recognition performance.

Motor imagery based brain-computer interfaces (MI-BCIs) have been widely used in intelligent medical applications such as post-stroke rehabilitation and mobile assistant robots. However, the high inter-subject variability and the non-stationarity of EEG characteristics limit the cross-subject applications of MI-BCIs. Long-term calibration can be used to improve EEG-based performance, but which will cause low efficiency and reduce practicality. To overcome the limitation, data from other subjects can be used for transfer learning to reduce calibration time. Therefore, a calibration-free transfer learning method for EEG-based cross-subject MI classification is proposed in this paper. On one hand, Euclidean alignment and Riemannian alignment are introduced to reduce domain differences. On the other hand, the similarity is calculated by Multiple Kernel-Maximum Mean Discrepancy (MK-MMD) to select appropriate source domain samples, which is followed by domain adversarial training of neural network (DANN) for the final model construction. In order to achieve calibration-free, the new subjects' resting-state data was used only. Extensive experiments were conducted on BCI competition IV dataset 2a. The results show that the proposed method can achieve 75.96% classification accuracy without using subjects' labeled data, which demonstrates the feasibility of the proposed method in calibration time reduction and classification accuracy improvement.

Emotion recognition plays an important part in human-computer interaction (HCI). Currently, the main challenge in electroencephalogram (EEG)-based emotion recognition is the non-stationarity of EEG signals, which causes performance of the trained model decreasing over time. In this paper, we propose a two-level domain adaptation neural network (TDANN) to construct a transfer model for EEG-based emotion recognition. Specifically, deep features from the topological graph, which preserve topological information from EEG signals, are extracted using a deep neural network. These features are then passed through TDANN for two-level domain confusion. The first level uses the maximum mean discrepancy (MMD) to reduce the distribution discrepancy of deep features between source domain and target domain, and the second uses the domain adversarial neural network (DANN) to force the deep features closer to their corresponding class centers. We evaluated the domain-transfer performance of the model on both our self-built data set and the public data set SEED. In the cross-day transfer experiment, the ability to accurately discriminate joy from other emotions was high: sadness (84%), anger (87.04%), and fear (85.32%) on the self-built data set. The accuracy reached 74.93% on the SEED data set. In the cross-subject transfer experiment, the ability to accurately discriminate joy from other emotions was equally high: sadness (83.79%), anger (84.13%), and fear (81.72%) on the self-built data set. The average accuracy reached 87.9% on the SEED data set, which was higher than WGAN-DA. The experimental results demonstrate that the proposed TDANN can effectively handle the domain transfer problem in EEG-based emotion recognition.

There is a correlation between adjacent channels of electroencephalogram (EEG), and how to represent this correlation is an issue that is currently being explored. In addition, due to inter-individual differences in EEG signals, this discrepancy results in new subjects need spend a amount of calibration time for EEG-based motor imagery brain-computer interface. In order to solve the above problems, we propose a Dynamic Domain Adaptation Based Deep Learning Network (DADL-Net). First, the EEG data is mapped to the three-dimensional geometric space and its temporal-spatial features are learned through the 3D convolution module, and then the spatial-channel attention mechanism is used to strengthen the features, and the final convolution module can further learn the spatial-temporal information of the features. Finally, to account for inter-subject and cross-sessions differences, we employ a dynamic domain-adaptive strategy, the distance between features is reduced by introducing a Maximum Mean Discrepancy loss function, and the classification layer is fine-tuned by using part of the target domain data. We verify the performance of the proposed method on BCI competition IV 2a and OpenBMI datasets. Under the intra-subject experiment, the accuracy rates of 70.42% and 73.91% were achieved on the OpenBMI and BCIC IV 2a datasets.

Deep learning based electroencephalography (EEG) signal processing methods are known to suffer from poor test-time generalization due to the changes in data distribution. This becomes a more challenging problem when privacy-preserving representation learning is of interest such as in clinical settings. To that end, we propose a multi-source learning architecture where we extract domain-invariant representations from dataset-specific private encoders. Our model utilizes a maximum-mean-discrepancy (MMD) based domain alignment approach to impose domain-invariance for encoded representations, which outperforms state-of-the-art approaches in EEG-based emotion classification. Furthermore, representations learned in our pipeline preserve domain privacy as dataset-specific private encoding alleviates the need for conventional, centralized EEG-based deep neural network training approaches with shared parameters.

Objective. Deep transfer learning has been widely used to address the nonstationarity of electroencephalogram (EEG) data during motor imagery (MI) classification. However, previous deep learning approaches suffer from limited classification accuracy because the temporal and spatial features cannot be effectively extracted. Approach. Here, we propose a novel end-to-end deep subject adaptation convolutional neural network (SACNN) to handle the problem of EEG-based MI classification. Our proposed model jointly optimizes three modules, i.e. a feature extractor, a classifier, and a subject adapter. Specifically, the feature extractor simultaneously extracts the temporal and spatial features from the raw EEG data using a parallel multiscale convolution network. In addition, we design a subject adapter to reduce the feature distribution shift between the source and target subjects by using the maximum mean discrepancy. By minimizing the classification loss and the distribution discrepancy, the model is able to extract the temporal-spatial features to the prediction of a new subject. Main results. Extensive experiments are carried out on three EEG-based MI datasets, i.e. brain–computer interface (BCI) competition IV dataset IIb, BCI competition III dataset IVa, and BCI competition IV dataset I, and the average accuracy reaches to 86.42%, 81.71% and 79.35% on the three datasets respectively. Furthermore, the statistical analysis also indicates the significant performance improvement of SACNN. Significance. This paper reveals the importance of the temporal-spatial features on EEG-based MI classification task. Our proposed SACNN model can make fully use of the temporal-spatial information to achieve the purpose.

Convolutional neural networks (CNNs) have become a powerful technique to decode EEG and have become the benchmark for motor imagery EEG Brain-Computer-Interface (BCI) decoding. However, it is still challenging to train CNNs on multiple subjects' EEG without decreasing individual performance. This is known as the negative transfer problem, i.e. learning from dissimilar distributions causes CNNs to misrepresent each of them instead of learning a richer representation. As a result, CNNs cannot directly use multiple subjects' EEG to enhance model performance directly. To address this problem, we extend deep transfer learning techniques to the EEG multi-subject training case. We propose a multi-branch deep transfer network, the Separate-Common-Separate Network (SCSN) based on splitting the network's feature extractors for individual subjects. We also explore the possibility of applying Maximum-mean discrepancy (MMD) to the SCSN (SCSN-MMD) to better align distributions of features from individual feature extractors. The proposed network is evaluated on the BCI Competition IV 2a dataset (BCICIV2a dataset) and our online recorded dataset. Results show that the proposed SCSN (81.8%, 53.2%) and SCSN-MMD (81.8%, 54.8%) outperformed the benchmark CNN (73.4%, 48.8%) on both datasets using multiple subjects. Our proposed networks show the potential to utilise larger multi-subject datasets to train an EEG decoder without being influenced by negative transfer.

Electroencephalogram (EEG) signals are not easily camouflaged, portable, and noninvasive. It is widely used in emotion recognition. However, due to the existence of individual differences, there will be certain differences in the data distribution of EEG signals in the same emotional state of different subjects. To obtain a model that performs well in classifying new subjects, traditional emotion recognition approaches need to collect a large number of labeled data of new subjects, which is often unrealistic. In this study, a transfer discriminative dictionary pair learning (TDDPL) approach is proposed for across-subject EEG emotion classification. The TDDPL approach projects data from different subjects into the domain-invariant subspace, and builds a transfer dictionary pair learning based on the maximum mean discrepancy (MMD) strategy. In the subspace, TDDPL learns shared synthesis and analysis dictionaries to build a bridge of discriminative knowledge from source domain (SD) to target domain (TD). By minimizing the reconstruction error and the inter-class separation term for each sub-dictionary, the learned synthesis dictionary is discriminative and the learned low-rank coding is sparse. Finally, a discriminative classifier in the TD is constructed on the classifier parameter, analysis dictionary and projection matrix, without the calculation of coding coefficients. The effectiveness of the TDDPL approach is verified on SEED and SEED IV datasets.

No abstract available

No abstract available

A cross-subject domain generalization (DG) approach with multiadversarial strategies (DGMA) is introduced to reduce brain-computer interfaces (BCIs) systems’ dependency on high-quality, subject-specific electroencephalographic (EEG) data, making it adaptable to unseen domains. DGMA leverages annotated training data from other subjects and consists of three modules: 1) prefeature extraction (PFE), enhancing EEG signal separability through preprocessing, data augmentation, and tangent space mapping; 2) distribution feature updater (DFU), aligning intersubject feature distributions with marginal maximum mean discrepancy (MMD); and 3) multiadversarial training (MAT), initially using gradient reversal layer (GRL) to amplify domain differences and classification loss, allowing the model to learn diverse domain-specific features before minimizing these differences to balance domain transferability and discriminability. DGMA is capable of better capturing domain-specific features while achieving stronger generalization compared with traditional methods focused solely on minimizing domain differences. Validated on four motor imagery datasets, DGMA achieved state-of-the-art accuracies of 76.1% on BCI Competition IV 2a and 72.4% on the 002-2014 dataset. Additional tests on a private fatigue dataset and the SEED dataset yielded accuracies of 99.5% and 86.6%, respectively. The code can be found at https://github.com/liuyici/DGMA-BCI

EEG signals exhibit non-stationary characteristics, particularly across different subjects, which presents significant challenges in the precise classification of mental workload levels when applying a trained model to new subjects. Domain adaptation techniques have shown effectiveness in enhancing the accuracy of cross-subject classification. However, current state-of-the-art methods for cross-subject mental workload classification primarily focus on global domain adaptation, which may lack fine-grained information and result in ambiguous classification boundaries. We proposed a novel approach called deep subdomain adaptation network with class confusion loss (DSAN-CCL) to enhance the performance of cross-subject mental workload classification. DSAN-CCL utilizes the local maximum mean discrepancy to align the feature distributions between the source domain and the target domain for each mental workload category. Moreover, the class confusion matrix was constructed by the product of the weighted class probabilities (class probabilities predicted by the label classifier) and the transpose of the class probabilities. The loss for maximizing diagonal elements and minimizing non-diagonal elements of the class confusion matrix was added to increase the credibility of pseudo-labels, thus improving the transfer performance. The proposed DSAN-CCL method was validated on two datasets, and the results indicate a significant improvement of 3∼10 percentage points compared to state-of-the-art domain adaptation methods. In addition, our proposed method is not dependent on a specific feature extractor. It can be replaced by any other feature extractor to fit new applications. This makes our approach universal to cross-domain classification problems.

The affective Brain-Computer Interface (aBCI) systems strive to enhance prediction accuracy for individual subjects by leveraging data from multiple subjects. However, significant differences in EEG (Electroencephalogram) feature patterns among subjects often hinder these systems from achieving the desired outcomes. Although studies have attempted to address this challenge using subject-specific classifier strategies, the scarcity of labeled data remains a major hurdle. In light of this, Domain Adaptation (DA) technology has gradually emerged as a prominent approach in the field of EEG-based emotion recognition, attracting widespread research interest. The crux of DA learning lies in resolving the issue of distribution mismatch between training and testing datasets, which has become a focal point of academic attention. Currently, mainstream DA methods primarily focus on mitigating domain distribution discrepancies by minimizing the Maximum Mean Discrepancy (MMD) or its variants. Nevertheless, the presence of noisy samples in datasets can lead to pronounced shifts in domain means, thereby impairing the adaptive performance of DA methods based on MMD and its variants in practical applications to some extent. Research has revealed that the traditional MMD metric can be transformed into a 1-center clustering problem, and the possibility clustering model is adept at mitigating noise interference during the data clustering process. Consequently, the conventional MMD metric can be further relaxed into a possibilistic clustering model. Therefore, we construct a distributed distance measure with Discriminative Possibilistic Clustering criterion (DPC), which aims to achieve two objectives: (1) ensuring the discriminative effectiveness of domain distribution alignment by finding a shared subspace that minimizes the overall distribution distance between domains while maximizing the semantic distribution distance according to the principle of “sames attract and opposites repel”; and (2) enhancing the robustness of distribution distance measure by introducing a fuzzy entropy regularization term. Theoretical analysis confirms that the proposed DPC is an upper bound of the existing MMD metric under certain conditions. Therefore, the MMD objective can be effectively optimized by minimizing the DPC. Finally, we propose a domain adaptation in Emotion recognition based on DPC (EDPC) that introduces a graph Laplacian matrix to preserve the geometric structural consistency between data within the source and target domains, thereby enhancing label propagation performance. Simultaneously, by maximizing the use of source domain discriminative information to minimize domain discrimination errors, the generalization performance of the DA model is further improved. Comparative experiments on several representative domain adaptation learning methods using multiple EEG datasets (i.e., SEED and SEED-IV) show that, in most cases, the proposed method exhibits better or comparable consistent generalization performance.

In EEG-based emotion recognition, traditional shallow convolutional methods often struggle with effective feature capture, particularly in cross-subject scenarios where domain alignment is crucial. This study proposes a novel approach: Utilizing a pre-trained model for rapid feature extraction, followed by the integration of Joint Maximum Mean Discrepancy(JMMD) loss within the Joint Adaptation Network(JAN) framework and dynamic routing in Capsule Networks, forming the dynamic domain adaptation model DA-PCN. The pre-trained VGG16 enhances the capture of complex emotional patterns and improves generalization across subjects. These features are further refined through a dual-layer Capsule Network, which aligns them across different domains at each layer. Assessments across various subjects and multiple videos from the same subjects using the DEAP dataset support the effectiveness of this method.

The affective Brain-Computer Interface (aBCI) systems, which achieve predictions for individual subjects through training on multiple subjects, often cannot achieve satisfactory results due to the differences in Electroencephalogram (EEG) patterns between subjects. One tried to use Subject-specific classifiers, but there was a lack of sufficient labeled data. To solve this problem, Domain Adaptation (DA) has recently received widespread attention in the field of EEG-based emotion recognition. Domain adaptation (DA) learning aims to solve the problem of inconsistent distributions between training and test datasets and has received extensive attention. Most existing methods use Maximum Mean Discrepancy (MMD) or its variants to minimize the problem of domain distribution inconsistency. However, noisy data in the domain can lead to significant drift in domain means, which can affect the adaptability performance of learning methods based on MMD and its variants to some extent. Therefore, we propose a robust domain adaptation learning method with possibilistic distribution distance measure. Firstly, the traditional MMD criterion is transformed into a novel possibilistic clustering model to weaken the influence of noisy data, thereby constructing a robust possibilistic distribution distance metric (P-DDM) criterion. Then the robust effectiveness of domain distribution alignment is further improved by a fuzzy entropy regularization term. The proposed P-DDM is in theory proved which be an upper bound of the traditional distribution distance measure method MMD criterion under certain conditions. Therefore, minimizing P-DDM can effectively optimize the MMD objective. Secondly, based on the P-DDM criterion, a robust domain adaptation classifier based on P-DDM (C-PDDM) is proposed, which adopts the Laplacian matrix to preserve the geometric consistency of instances in the source domain and target domain for improving the label propagation performance. At the same time, by maximizing the use of source domain discriminative information to minimize domain discrimination error, the generalization performance of the learning model is further improved. Finally, a large number of experiments and analyses on multiple EEG datasets (i.e., SEED and SEED-IV) show that the proposed method has superior or comparable robustness performance (i.e., has increased by around 10%) in most cases.

No abstract available

Because of the non-steady state of EEG signals, there are differences in the distribution of electroencephalogram (EEG) data among different subjects. This distribution difference leads to a large error indirectly using the data of different subjects for training. This paper proposes a cross-subject trial reweighting (CSTR) method to reduce the distribution difference. CSTR assigns weights to each sample to narrow the maximum mean discrepancy between the source and target domains. CSTR is applied to the original EEG data samples, and can also be applied to samples after various feature processing. We use the motor imagery dataset to verify the effectiveness of the CSTR algorithm. The experimental results show that the source domain and the target domain become more similar after the trials are reweighted. The classification performance is improved using the reweighted data. The method proposed in this paper can improve the performance of the transfer learning brain-computer interface, reduce the calibration time of BCI, and promote the practical application of BCI.

Maximum Mean Discrepancy (MMD) is a widely used concept in machine learning research which has gained popularity in recent years as a highly effective tool for comparing (finite-dimensional) distributions. Since it is designed as a kernel-based method, the MMD can be extended to path space valued distributions using the signature kernel. The resulting signature MMD (sig-MMD) can be used to define a metric between distributions on path space. Similarly to the original use case of the MMD as a test statistic within a two-sample testing framework, the sig-MMD can be applied to determine if two sets of paths are drawn from the same stochastic process. This work is dedicated to understanding the possibilities and challenges associated with applying the sig-MMD as a statistical tool in practice. We introduce and explain the sig-MMD, and provide easily accessible and verifiable examples for its practical use. We present examples that can lead to Type 2 errors in the hypothesis test, falsely indicating that samples have been drawn from the same underlying process (which generally occurs in a limited data setting). We then present techniques to mitigate the occurrence of this type of error.

Existing two-sample testing techniques, particularly those based on choosing a kernel for the Maximum Mean Discrepancy (MMD), often assume equal sample sizes from the two distributions. Applying these methods in practice can require discarding valuable data, unnecessarily reducing test power. We address this long-standing limitation by extending the theory of generalized U-statistics and applying it to the usual MMD estimator, resulting in new characterization of the asymptotic distributions of the MMD estimator with unequal sample sizes (particularly outside the proportional regimes required by previous partial results). This generalization also provides a new criterion for optimizing the power of an MMD test with unequal sample sizes. Our approach preserves all available data, enhancing test accuracy and applicability in realistic settings. Along the way, we give much cleaner characterizations of the variance of MMD estimators, revealing something that might be surprising to those in the area: while zero MMD implies a degenerate estimator, it is sometimes possible to have a degenerate estimator with nonzero MMD as well; we give a construction and a proof that it does not happen in common situations.

In electroencephalogram (EEG)-based emotion recognition, the applicability of most current models is limited by inter-subject variability and emotion complexity. This study proposes a multi-task adversarial domain adaptation (MTADA) network to enhance cross-subject emotion recognition performance. The model first employs a domain matching strategy to select the source domain that best matches the target domain. Then, adversarial domain adaptation is used to learn the difference between source and target domains, and a fine-grained joint domain discriminator is constructed to align them by incorporating category information. At the same time, a multi-task learning mechanism is utilized to learn the intrinsic relationships between different emotions and predict multiple emotions simultaneously. We conducted comprehensive experiments on two public datasets, DEAP and FACED. On DEAP, the average accuracies for valence, arousal and dominance are 76.39%, 69.74% and 68.26%, respectively. On FACED, the average accuracies for valence and arousal are 78.90% and 77.95%. When using the subject from DEAP as the source domain to predict the subjects in FACED, the accuracies for valence and arousal are 61.07% and 60.82%. These results show that our MTADA model improves cross-subject emotion recognition and outperforms most state-of-the-art methods, which may provide new approach for EEG-based emotion brain-computer interface systems.

No abstract available

In recent years, unsupervised domain adaptation (UDA) has emerged as a promising approach for constructing cross-subject emotion recognition models. However, most existing UDA methods do not fully exploit class information in the target domain, resulting in a relatively coarse-grained process of domain alignment and a higher risk of incorrect class matching. To address this issue, this paper proposes a novel adversarial training-based domain adaptation framework. The proposed method leverages emotion class prototypes to enhance intra-class correlation between the source and target feature distributions. Meanwhile, soft pseudo-labels generated by prototype clustering are utilized to further improve the inter-class discriminability within each domain. In order to enhance the robustness and quality of hard pseudo-labels in the target domain, a dual pseudo-labeling strategy is introduced. Finally, adversarial training is conducted to achieve a more fine-grained alignment of data distributions across domains. We conduct cross-subject and cross-session evaluations on the SEED and SEED-IV datasets, respectively. Experimental results demonstrate the effectiveness of our method and its advantages over several state-of-the-art UDA approaches. By introducing dual pseudo-labels, our study incorporates additional supervision, enabling a more refined domain adaptation process and significantly improving the generalization capability of EEG-based emotion recognition models.

No abstract available

The upcoming era of wearable health monitoring devices has created a need for automated signal processing algorithms that can be trained with a minimal amount of labeled data. In our previous work, we showed that transfer learning techniques like semi-supervised adversarial domain adaptation can help to achieve this. We applied our method to remote sleep monitoring, by performing sleep staging on single-channel wearable EEG signals. In this work, we propose data augmentation to help in tackling this challenge. By using an artificially increased amount of labeled data, our semi-supervised adversarial domain adaptation method improves its performance on the wearable EEG data. The accuracy is increased consistently by 0.6% to 1.4% relative to the results without augmentation. As both adversarial domain adaptation and data augmentation are strategies to deal with the scarceness of data, we conclude that these methods are can effectively be combined to surpass their individual performance.

Sleep-stage classification is a critical aspect of understanding sleep patterns in sleep research and healthcare. However, challenges arise when dealing with a limited number of labeled samples in the target domain. Traditional methods in Deep Learning (DL) and Domain Adaptation (DA) globally compare feature distributions, often overlooking intricate decision boundaries between sleep-stage classes. This results in ambiguous features near class boundaries, diminishing classification accuracy. The conventional two-step process of using a pre-trained classifier for predictions and assessing uncertainty fails to effectively incorporate unlabeled data in classifier training, neglecting the complexities of the target domain. To address these challenges, we propose Adversarial Deep Learning Joint Domain Adaptation (ADLJDA). This innovative approach integrates an adversarial model and deploys two distinct sleep-stage classifiers as discriminators, allowing for a nuanced consideration of class boundaries during feature distribution alignment. ADLJDA also incorporates an entropy measure with cross-entropy loss during training to harness information from unlabeled data in the target domain. Experimental results on three benchmark EEG datasets highlight the efficacy of ADLJDA. The approach consistently demonstrates the ability to generate robust and transferable features, mitigating the impact of ambiguous features near original class boundaries. Importantly, ADLJDA shows a significant improvement in classification accuracy compared to existing state-of-the-art DA methods, even in datasets with intricate patterns and complexities. This research contributes to advancing sleep-stage classification methodologies, offering a promising solution for enhanced accuracy in real-world applications and furthering our understanding of sleep-related phenomena.

Decoding the human emotional states based on electroencephalography (EEG) in affective brain-computer interfaces (BCI) is a great challenge due to inter-subject variability. Existing methods mostly use large amounts of EEG data of each new subject to calibrate the algorithm, which could be time-consuming and not user-oriented. To address this issue, we propose a combination of using transformers (TF) and adversarial discriminative domain adaptation (ADDA) to perform the emotion recognition task in a cross-subject manner. TF principally relies on the attention mechanism. Our proposed approach performs scaledot product attention on the feature-channel aspect of EEG data to improve the spatial features. Then, the temporal transforming is applied to get the global discriminative representations from the time component. Moreover, ADDA aims to minimize the discrepancy of EEG data from various subjects. We evaluate the proposed ADDA-TF on the publicly available DEAP dataset and demonstrate the improvements it provides on low versus high valence and arousal classification.

No abstract available

Cross-subject emotion recognition is one of the most challenging tasks in electroencephalogram (EEG)-based emotion recognition. To guarantee the constancy of feature representations across domains and to eliminate differences between domains, we explored the feasibility of combining temporal convolutional networks (TCNs) and adversarial discriminative domain adaptation (ADDA) algorithms in solving the problem of domain shift in EEG-based cross-subject emotion recognition. In light of EEG signals that have specific temporal properties, we chose the temporal model TCN as the feature encoder. To verify the validity of the proposed method, we conducted experiments on two public datasets: DEAP and DREAMER. The experimental results show that for the leave-one-subject-out evaluation, average accuracies of 64.33% (valence) and 63.25% (arousal) were obtained on the DEAP dataset, and average accuracies of 66.56% (valence) and 63.69% (arousal) were achieved on the DREAMER dataset. Extensive experiments demonstrate that our method for EEG-based cross-subject emotion recognition is effective.

To address the problem that some source domain samples that are difficult to transfer disturb the target domain data distribution due to the difference in transfer value between different motor imagery electroencephalogram (MI-EEG) sample data, and that the model has poor feature extraction and classification performance when adapting to different motor imagery datasets, this paper improves the conditional domain adversarial network (CDAN) method introduced by domain generalization technology, and proposes a conditional domain adaptation network based on sample weight (SW-CDAN) method. This method makes the entropy output by the domain discriminator as the sample weight, which is used to adjust the classification loss during the model training process, so that the model can extract transferable features from the common features of the data, thereby enhancing the model’s category prediction ability and model generalization ability. The experimental results show that the SW-CDAN method can effectively improve the classification performance and model generalization ability of motor imagery EEG signals, so that even when facing a small amount of motor imagery EEG signals with low effective components, it can still maintain a high classification accuracy. The SWCDAN method achieves relatively high classification accuracy on BCI Competition IV 2a dataset, which is about 1.87% higher than CDAN method respectively.

Electroencephalography (EEG) is the most prevalent signal acquisition technique for brain-computer interface (BCI). However, the statistical distribution of EEG data varies across subjects and sessions, resulting in poor generalization of the domain-specific classifier. Although the collection of a large number of recordings may alleviate this issue, it is often impractical and not user-friendly. This study proposes the integration of deep domain adaptation with few-shot learning to address the challenge by leveraging the knowledge from multiple source subjects to enhance the performance of a single target subject. The framework incorporated 3 modules: a feature extractor, domain discriminator, and classifier. The feature extractor utilized the available labeled samples with supervised contrastive loss to map the discriminate features onto a deep representation space, where the features from the same class were more similar than those from different classes. The domain discriminator was used to reduce domain drift, through adversarial training. The classifier predicted the user motor intention, based on EEG features. The framework was extensively evaluated through the BCI Competition IV Datasets 2a and 2b. The results of this study indicate that the framework is capable of enhancing the BCI performance and potentially decreases the calibration effort compared to the traditional approach, but the major limitation of this method is that it requires meticulous selection of source subjects.

Alzheimer’s Disease (AD) and Mild Cognitive Impairment (MCI) are neurodegenerative disorders that require early and accurate diagnosis for effective intervention. Electroencephalography (EEG) is a non-invasive tool for detecting cognitive decline, but subject variability poses a significant challenge in classification models. This work proposes Neurological Domain Adaptation with Transformer (NDAT), a multi-input Transformer-based framework that incorporates Instance Normalization (IN) and Adversarial Domain Adaptation (ADA) for subject-independent EEG-based classification of AD and MCI. The model extracts feature from 1D EEG signals using a Transformer encoder and from 2D EEG spectrograms using a Custom Convolutional Neural Network (Custom CNN). A fusion network aligns these multi-modal features for final classification. To mitigate subject-specific biases, Instance normalization is applied to the extracted features. Additionally, ADA is integrated using a Gradient Reversal Layer (GRL), ensuring the model learns domain-invariant representations for robust subject-independent classification. The framework is evaluated on two EEG datasets: one for Alzheimer’s disease classification (Normal, Frontotemporal Dementia (FTD), AD) and another for MCI classification (Normal, MCI, AD). To address the class imbalance in the FTD category, augmentation, and resampling techniques are applied to improve generalization. Experimental results demonstrate that NDAT significantly outperforms conventional methods, achieving high accuracy, sensitivity, and specificity in both subject-dependent and subject-independent settings. These findings highlight the effectiveness of deep learning-based feature extraction, domain adaptation, and normalization strategies in enhancing EEG-based neurodegenerative disease classification.

EEG signals contain valuable physiological and psychological information, essential for brain-computer interfaces (BCIs) and neurorehabilitation. However, the non-stationarity and subject-specific variability of EEG make cross-subject generalization difficult, limiting practical deployment without costly recalibration. Unsupervised domain adaptation (UDA) aims to reduce domain discrepancies for better generalization. While many UDA methods use discrepancy-based or adversarial strategies to extract domain-invariant features, they are limited by Euclidean space, which doesn't capture EEG's nonlinear relationships. To address this, we propose a novel UDA framework on the Riemannian manifold of SPD matrices, using adversarial training. Our approach features a dual flow architecture with separate feature extractors for the source and target domains, incorporating a manifold soft parameter sharing mechanism and multi-level alignment loss for better domain alignment and feature separability. Unlike traditional methods, our model preserves domain-specific structures while aligning the feature extractors' outputs, capturing more effective domain-invariant information. Experiments on three BCI datasets show that our method outperforms several state-of-the-art UDA approaches in cross-subject EEG classification.

To mitigate the negative transfer impact from different domains in motor imagery electroencephalogram (MI-EEG) based brain computer interface, this paper proposes a Domain Adaptation Method with Domain Selection in brain source space (DADSDSS) to enhance the spatial-temporal-geometric feature similarity across subjects. Neural electrical activity is estimated through EEG source imaging techniques to calculate regional equivalent dipoles. Inter-domain similarity is then assessed by using Canonical Correlation Analysis (CCA)-based K-means for centroid generation and correlation measurement, and the Kuhn–Munkres (KM) algorithm for optimal class matching. A Graph Convolutional block (GCB)-based spatio-temporal feature extractor and a Riemannian space feature extractor are designed to comprehensively capture the spatial-temporal and high-order geometric features successively. Futhermore, two discriminators are applied to learn the domain-invariant features through adversarial training and classify MI tasks. The experiments were conducted based on a public MI-EEG dataset with 4-MI tasks, DADSDSS achived the relatively higher average recognition accuracy (81.29%), Kappa value (0.7505), and F1-score (0.8159), respectively. The results indicate it is effective to reduce the difference from source and target domains via domain selection in the brain source space, which is matched well with the unique feature extractors and discriminators, yielding better domain adaptation performance.

To build a subject-independent affective model based on electroencephalography (EEG) is a challenging task due to the domain shift problem caused by individual differences in EEG data. In this paper, we prove a new generalization bound based on Wasserstein distance for multi-source classification and regression problems. Based on our bound, we propose two novel Wasserstein-distance-based multi-source adversarial domain adaptation methods (wMADA) for learning domain invariant and task discriminative domain mappings by dynamically aligning different domain mappings. We evaluate our methods on two typical EEG datasets. The experimental results demonstrate that our wMADA methods successfully handle the multi-source domain shift problem in creating subject-independent affective models and outperform the state-of-the-art domain adaptation methods.

The identification of electroencephalography (EEG) signals’ cross sessions and subjects remains challenging due to the variability of data caused by extraneous factors and individual differences in EEG signals. Existing domain-adaptive transfer methods using cross-domain labeled samples for classification are too coarse and could lead to negative transfer problems. To solve this problem, we propose a prototypical contrastive domain adaptation (PCDA) network in this article. First, we align the data from different domains to reduce the data distribution differences for supporting the subsequent model construction. Then, a conditional domain adversarial network is used in the feature extraction stage to achieve domain alignment and learn deep feature representations. Second, we propose a scoring method to equivalently quantify the similarity of data from different domains using resting-state data and select similar source domain data to fine-tune the model. Finally, we propose a prototypical contrastive (PC) learning module. In-domain PC learning captures and compares the category-wise semantic structure of the data and the learned representations to enable the clustering of similar features. Cross-domain PC learning encodes and compares the semantic structure in shared embedding space to enable self-supervised feature alignment and reduce negative transfer. The experimental results show that the PCDA network achieves better results on the datasets of brain-computer interface (BCI) Competition IV II-a and II-b, and the ablation experiments validate the efficacy of the method.

Motor imagery (MI) based Brain-computer interface (BCI) is a promising BCI paradigm that can help neuromuscular injury patients to recover or replace their motor abilities. However, electroencephalography (EEG) based MI-BCI suffers from its long calibration time and low classification accuracy, which restrict its application. Thus, it is important to reduce the calibration time of MI-BCI and enhance its prediction accuracy. In this study, we propose a filter bank Wasserstein adversarial domain adaptation framework (FBWADA) that uses a short amount of training data from a new target subject, and all collected data from an existing subject. A Convolutional Neural Networks (CNN) based feature extractor is designed to extract feature from EEG data. Filter bank strategy is employed to extract feature from multiple sub bands and integrate predictions from all sub bands. Wasserstein Generative Adversarial Networks (WGAN) based domain adaptation network aligns the marginal and conditional distribution of target and source. We evaluate our method on Data set 2a of BCI competition IV. Experiment results show that our method achieves the best performance among compared methods under different amount of training data. Performance of our method trained with certain blocks of data is similar to or better than the best comparing method trained with one more block. This indicates that our method could reduce the need for training data for at least one block.

EEG signals have emerged as a potent modality for emotion recognition owing to their direct correspondence to brain activity. However, the considerable inter-individual variability of EEG signals and the intricacy of their spatial structural features pose significant challenges. To address these challenges, we introduce the Multi-View Hierarchical Attention Graph Convolutional Network (MAGCN) accompanied by Domain Adaptation, a novel architecture that effectively combines multi-view adjacency matrices with a hierarchical attention mechanism and domain adaptation for robust EEG-based emotion recognition. The MAGCN model leverages both functional and spatial connections within EEG data, facilitating dynamic graph convolution that readily adapts to the variability of emotional states. Our domain discriminator, inspired by adversarial learning, guarantees the extraction of domain-invariant features, subsequently augmenting the model's generalisation across subjects. Experiments on the SEED and SEED-IV datasets attest to the MAGCN model's superior performance over existing methods.

No abstract available

Traditional electroencephalograph (EEG)-based emotion recognition requires a large number of calibration samples to build a model for a specific subject, which restricts the application of the affective brain computer interface (BCI) in practice. We attempt to use the multi-modal data from the past session to realize emotion recognition in the case of a small amount of calibration samples. To solve this problem, we propose a multi-modal domain adaptive variational autoencoder (MMDA-VAE) method, which learns shared cross-domain latent representations of the multi-modal data. Our method builds a multi-modal variational autoencoder (MVAE) to project the data of multiple modalities into a common space. Through adversarial learning and cycle-consistency regularization, our method can reduce the distribution difference of each domain on the shared latent representation layer and realize the transfer of knowledge. Extensive experiments are conducted on two public datasets, SEED and SEED-IV, and the results show the superiority of our proposed method. Our work can effectively improve the performance of emotion recognition with a small amount of labelled multi-modal data.