食品感官基于脑电(EEG)的有监督学习：味觉、气味、风味感知、喜好、愉悦度、接受度、食欲及消费者偏好的分类、预测或解码

Accurate detection of human emotion is an important topic for affective computing. Especially with the rise of artificial intelligence in the marketing industry, the tools available are subjective and often heavily dependent on sample sizes and demographics. This study explores the neural responses to olfactory stimuli by analyzing EEG data collected from 57 participants exposed to a perfume scent in correlation with self-reported survey results. The electroencephalogram (EEG) signals were processed to extract time-domain, spectral-domain, and nonlinear features, which were subsequently classified using various machine learning algorithms. The classification outcomes were mapped onto a two-dimensional pleasure-arousal plane, with the Medium Gaussian support vector machine (SVM) achieving the highest performance, including $99.8 \%$ validation accuracy and $100 \%$ test accuracy. These results highlight the significant potential of EEG-based approaches in decoding the neural underpinnings of sensory experiences, with implications for applications in neuromarketing and therapeutic contexts.

The human olfactory system's temporal dynamics are crucial for sensory perception. By learning the temporal dynamics of EEG and utilizing breathing signals, we aim to better understand the neural features of olfactory perception from EEG. To decode the olfactory response effectively, we introduce a new method: the Token Alignment and Cross-Attention Fusion network (TACAF), a multimodal deep learning framework that enhances olfactory EEG decoding using wavelet features for time window selection and spectral analysis for data representation. Spatial features are extracted using spatial learning modules, and temporal dynamics are captured through a multi-head self-attention mechanism. The Temporal Token Semantic Alignment (TTSA) module synchronizes breathing information with EEG data for effective fusion. We collected EEG recordings and breathing signals from 20 subjects to study the decoding responses to pleasant and unpleasant odors. Our evaluation shows that TACAF significantly outperforms existing methods. Further analysis indicates that prolonged odor exposure leads to olfactory adaptation, reducing recognition performance. The findings are visualized through spatial topology maps with saliency mappings, providing insights into the neural mechanisms of olfactory perception.

BACKGROUND Decoding olfactory-induced electroencephalography (olfactory EEG) signals has gained significant attention in recent years, owing to its potential applications in several fields, such as disease diagnosis, multimedia applications, and brain-computer interaction (BCI). Extracting discriminative features from olfactory EEG signals with low spatial resolution and poor signal-to-noise ratio is vital but challenging for improving decoding accuracy. NEW METHODS By combining discrete wavelet transform (DWT) with one-versus-rest common spatial pattern (OVR-CSP), we develop a novel feature, named wavelet-spatial domain feature (WSDF), to decode the olfactory EEG signals. First, DWT is employed on EEG signals for multilevel wavelet decomposition. Next, the DWT coefficients obtained at a specific level are subjected to OVR-CSP for spatial filtering. Correspondingly, the variance is extracted to generate a discriminative feature set, labeled as WSDF. RESULTS To verify the effectiveness of WSDF, a classification of olfactory EEG signals was conducted on two data sets, i.e., a public EEG dataset 'Odor Pleasantness Perception Dataset (OPPD)', and a self-collected dataset, by using support vector machine (SVM) trained based on different cross-validation methods. Experimental results showed that on OPPD dataset, the proposed method achieved a best average accuracy of 100% and 94.47% for the eyes-open and eyes-closed conditions, respectively. Moreover, on our own dataset, the proposed method gave a highest average accuracy of 99.50%. COMPARISON WITH EXISTING METHODS Compared with a wide range of EEG features and existing works on the same dataset, our WSDF yielded superior classification performance. CONCLUSIONS The proposed WSDF is a promising candidate for decoding olfactory EEG signals.

Detection of olfactory stimulus in electroencephalogram

BACKGROUND While decoding visual and auditory stimuli using recorded EEG signals has enjoyed significant attention in the past decades, decoding olfactory sensory input from EEG data remains a novelty. Recent interest in the brain's mechanisms of processing olfactory stimuli partly stems from the association of the olfactory system and its deficit with neurodegenerative diseases. NEW METHODS An olfactory stimulus decoder using features that represent nonlinear behavior content in the recorded EEG data has been introduced for classifying 4 olfactory stimuli in 5 healthy male subjects. RESULTS We show that by using nonlinear and chaotic features, a subject-specific classifier can be developed for identifying the odors that subjects perceive with an average accuracy of 96.71% and 88.79% in the eyes-open and eyes-closed conditions, respectively. We also employ our methodology in building cross-subject classifiers: once for identifying pleasant and unpleasant odors, and once for the classification of all four olfactory stimuli. The accuracy of our proposed methodology is 91.7% and 82.1% in the eyes-open and eyes-closed conditions, for the odor pleasantness classification. The accuracy of cross-subject classification of all odors is 64.3% and 54.8% for the eyes-open and eyes-closed conditions, respectively, which is well above chance level. COMPARISON WITH EXISTING METHODS Comparison with similar studies reveals that our proposed method outperforms other classification schemes in terms of accuracy. CONCLUSIONS The results can help researchers design more accurate classifiers for the detection of perceived odors using EEG signals. These results can contribute to gaining more insight into the brain's process of odor perception.

Significance To elucidate when and where in the brain different aspects of odor perception emerge, we decoded odors from an electroencephalogram and associated the results with perception and source activities. The odor information was decoded 100 ms after odor onset at the earliest, with its signal sources estimated in and around the olfactory areas. The neural representation of odor unpleasantness emerged 300 ms after odor onset, followed by pleasantness and perceived quality at 500 ms. During this time, brain regions representing odor information spread rapidly from the olfactory areas to regions associated with emotional, semantic, and memory processing. The results suggested that odor perception emerges through computations in these areas, with different perceptual aspects having different spatiotemporal dynamics.

… -subject olfactory EEG recognition dramatically limits its application. In this paper, a human–machine collaborative multimodal learning method is proposed for cross-subject olfactory …

Olfactory stimuli offer a unique and underexplored modality for affective brain-computer interfaces (aBCIs), leveraging the direct projection of odors to limbic structures involved in emotion processing. In this paper, we introduce a novel emotion induction paradigm based on olfactory stimulation and present a new dataset called SEED-OLF, which is designed to investigate the neural patterns of odor-induced emotional states. SEEDOLF was collected from three emotionrelated classification tasks: 1) distinguishing sniffing from non-sniffing EEG responses, 2) decoding subjectively reported valence, and 3) decoding the objective valence of odors. To advance the decoding of odor-induced emotional states, we introduce DSEN, a dual-stream EEG network that jointly captures spatiotemporal and spectral characteristics of EEG signals. DSEN integrates a spatiotemporal stream that employs temporal attention, multi-scale dense feature extraction, and attention-based fusion modules, along with a spectral stream that leverages differential entropy features and Transformer-based spectral processing. Experimental results indicate that DSEN significantly outperforms the existing state-of-the-art models across all three classification tasks. Topographic EEG analyses further reveal spectral-specific and region-specific cortical activations associated with olfactory-induced emotional processing. The SEED-OLF dataset and codes will be public upon paper acceptance.

Objective This study aimed to investigate whether electroencephalogram (EEG) data can be used to classify normosmia and anosmia using machine learning approaches and to examine how classification performance varies across single and combined chemical stimulation conditions. Methods EEG signals were recorded from 66 participants (26 normosmic and 40 anosmic) during olfactory stimulation with phenyl ethyl alcohol (PEA), hydrogen sulfide (H₂S), and carbon dioxide (CO₂), presented to the left and right nostrils. EEG data were averaged for each condition, resulting in six chemosensory event-related potential (CSERP) datasets per participant. Three machine learning models—support vector machine (SVM), random forest (RF), and convolutional neural network (CNN)—were used to classify normosmia and anosmia. Results Among all single and combined odor conditions, the combination of PEA_Right + H₂S_Left + H₂S_Right achieved the highest performance (ACC = 86.37%, F1 = 0.88, AUC = 0.85). CNN generally outperformed SVM and RF across most conditions. H₂S stimulation consistently provided strong discriminative power, particularly under left-sided presentation. Conclusion Machine learning applied to EEG-based CSERPs can effectively distinguish olfactory dysfunction, demonstrating the feasibility of this approach for objective evaluation of olfactory processing. Significance These findings suggest that EEG-based machine learning can serve as an objective framework for assessing olfactory function and may contribute to future development of data-driven diagnostic strategies for smell disorders.

… in deep learning models can effectively avoid the interference of redundant information in olfactory EEG … In this study, we selected 8 food odors to evoke the olfactory EEG responses of …

… to boost EEG decoding accuracy which lets researchers create EEG decoding systems that … scientists to build advanced deep learning models for analyzing olfactory EEG data in their …

Brain-Computer Interfaces (BCIs) are devices designed for establishing communication between the central nervous system and a computer. The communication can occur through different sensory modalities, and most commonly visual and auditory modalities are used. Here we propose that BCIs can be expanded by the incorporation of olfaction and discuss the potential applications of such olfactory BCIs. To substantiate this idea, we present results from two olfactory tasks: one that required attentive perception of odors without any overt report, and the second one where participants discriminated consecutively presented odors. In these experiments, EEG recordings were conducted in healthy participants while they performed the tasks guided by computer-generated verbal instructions. We emphasize the importance of relating EEG modulations to the breath cycle to improve the performance of an olfactory-based BCI. Furthermore, theta-activity could be used for olfactory-BCI decoding. In our experiments, we observed modulations of theta activity over the frontal EEG leads approximately 2 s after the inhalation of an odor. Overall, frontal theta rhythms and other types of EEG activity could be incorporated in the olfactory-based BCIs which utilize odors either as inputs or outputs. These BCIs could improve olfactory training required for conditions like anosmia and hyposmia, and mild cognitive impairment.

The final aim of our research is to develop a braincomputer interface (BCI) for olfaction. Our research program relies on modern olfactory displays and advanced processing of electroencephalography (EEG) and respiratory data in order to develop methods for robust olfactory BCI systems. Here we present the initial results from 17 subjects of our ongoing study. Applying k-nearest neighbors algorithm (k-NN) classification methods we achieve up to 79.9% accuracy for within subject classification of EEG signals for odor pairs. We propose some methods for further improvement of classification algorithm. EEG classification for different olfactory stimuli is the first step in this research, followed by the development of odor-imagery BCIs and odor-based neurofeedback.

In this study, a method is proposed to detect the presence of olfactory stimuli from Electroencephalogram (EEG) signals to be used in neuromarketing applications. Odor is used in different ways in neuromarketing applications since it stimulates various emotions. Multi-channel EEG signals were recorded from the volunteers while they were subjected to two open boxes of unscented and scented products in succession. After the necessary preprocessing steps, EEG sub-band powers were calculated for 14 EEG channels. These features were classified using machine learning methods, and the EEG segments in which the olfactory stimulus was present were classified. The results show that the proposed method gives successful results with 92% accuracy, 93% precision, 92% recall, and 92% F1-score using the Random Forest classifier.

In this study, we explore the feasibility of single-trial predictions of odor registration in the brain using olfactory bio-signals. We focus on two main aspects: input data modality and the processing model. For the first time, we assess the predictability of odor registration from novel electrobulbogram (EBG) recordings, both in sensor and source space, and compare these with commonly used electroencephalogram (EEG) signals. Despite having fewer data channels, EBG shows comparable performance to EEG. We also examine whether breathing patterns contain relevant information for this task. By comparing a logistic regression classifier, which requires hand-crafted features, with an end-to-end convolutional deep neural network, we find that end-to-end approaches can be as effective as classic methods. However, due to the high dimensionality of the data, the current dataset is insufficient for either classifier to robustly differentiate odor and non-odor trials. Finally, we identify key challenges in olfactory BCIs and suggest future directions for improving odor detection systems.

The sense of smell, which is also known as olfaction, can improve brain-computer interfaces (BCIs). It provides a natural and non-invasive way for users to interact with technology by assigning different commands to various scents delivered one after the other in a classical oddball paradigm setting. Olfactory BCIs detect changes in brain activity patterns in response to odors, which can be used to control devices, communicate, or provide information about the user’s mental state in the passive BCI modality. Olfactory stimulants can be processed quickly without causing attention overload, making them a promising direction for future BCI research and development. However, some challenges need to be overcome, such as the need for more accurate and reliable odor delivery systems and the development of robust algorithms for detecting and interpreting brain activity patterns. In a pilot study, we have shown the possibility of using a common spatial pattern (CSP) filtration and subsequent clustering of attended versus ignored scent stimuli in a novel BCI modality. Our preliminary results are promising, with accurate EEG response classification observed in four of eight experimental subjects.

Collecting emotional physiological signals is significant in building affective Human-Computer Interactions (HCI). However, how to evoke subjects’ emotions efficiently in EEG-related emotional experiments is still a challenge. In this work, we developed a novel experimental paradigm that allows odors dynamically participate in different stages of video-evoked emotions, to investigate the efficiency of olfactory-enhanced videos in inducing subjects’ emotions; According to the period that the odors participated in, the stimuli were divided into four patterns, i.e., the olfactory-enhanced video in early/later stimulus periods (OVEP/OVLP), and the traditional videos in early/later stimulus periods (TVEP/TVLP). The differential entropy (DE) feature and four classifiers were employed to test the efficiency of emotion recognition. The best average accuracies of the OVEP, OVLP, TVEP, and TVLP were 50.54%, 51.49%, 40.22%, and 57.55%, respectively. The experimental results indicated that the OVEP significantly outperformed the TVEP on classification performance, while there was no significant difference between the OVLP and TVLP. Besides, olfactory-enhanced videos achieved higher efficiency in evoking negative emotions than traditional videos. Moreover, we found that the neural patterns in response to emotions under different stimulus methods were stable, and for Fp1, FP2, and F7, there existed significant differences in whether adopt the odors.

Olfaction is crucial to our dietary choices and significantly influences our emotional and cognitive landscapes. Understanding the underlying neural mechanisms is pivotal, especially through the use of electroencephalography (EEG). This technology has strong temporal resolution, allowing it to capture the dynamics of neural responses to odors, bypassing the need for subjective interpretations. The application of EEG in food flavor research is still relatively new, but it has great potential. This review begins with an examination of general scent stimulation, charts the advances in using EEG to understand odor perception, and explores its future in food flavor science. By analyzing EEG's ability to detect distinct patterns and strengths in brain activity, we can elucidate the perceptual, affective, and cognitive frameworks associated with food odors. Event-related potentials and oscillatory activities, markers of central olfactory processing, provide insights into the neural architecture of olfaction. These markers are instrumental in assessing the influence of food odors on health, emotions, and decision-making processes. We argue that EEG's application in olfaction research holds considerable promise for the food industry to innovate products that are not only healthier but also more appealing, thereby promoting human well-being.

… Riemannian geometry classification framework to select the optimal channel sets for olfactory EEG … automatically identifying odor-induced pleasantness, a binary classification strategy …

Flavor-induced sensory satisfaction is critical for food acceptance and market success. However, traditional sensory evaluation methods, relying heavily on subjective assessments, often fail to accurately reflect real-time, objective neural processing underlying complex multisensory flavor experiences. This limitation highlights the need for innovative methods that objectively quantify how flavors are perceived and integrated within the brain. In this review, we first examine the neural pathways underlying flavor perception, focusing on how gustatory, olfactory, and oral somatosensory inputs interact with reward and hedonic networks to form integrated flavor experience. Building on this foundation, we then outline the latest strategies for developing flavor-oriented brain-computer interface (flavor-BCI), summarizing key features of various neuroimaging techniques and associated technical implementation workflows. Finally, we assess emerging applications of flavor-BCI in sensory assessment and consumer decision-making and identify opportunities and challenges for future food design and product development. Flavor perception begins with parallel encoding of chemical stimuli in the primary gustatory and olfactory cortices and in trigeminal pathways. These signals are subsequently integrated in higher order regions, forming a distributed neural network across cortical, limbic, and subcortical structures that support flavor recognition, hedonic appraisal, and motivated eating. Flavor-BCI systems record neural activity from these regions using electrophysiology or neuroimaging and apply advanced algorithms to decode neural representations, translating them into objective sensory outputs. Relative to traditional evaluations, this approach enables real-time, precise quantification of flavor experience. Flavor-BCI thus offers promising avenues for intelligent sensory evaluation and novel human-machine interactions.

Neuromarketing exploits neuroimaging techniques so as to reinforce the predictive power of conventional marketing tools, like questionnaires and focus groups. Electroencephalography (EEG) is the most commonly encountered neuroimaging technique due to its non-invasiveness, low-cost, and its very recent embedding in wearable devices. The transcription of brainwave patterns to consumer attitude is supported by various signal descriptors, while the quest for profitable novel ways is still an open research question. Here, we suggest the use of sample covariance matrices as alternative descriptors, that encapsulate the coordinated neural activity from distinct brain areas, and the adoption of Riemannian geometry for their handling. We first establish the suitability of Riemannian approach for neuromarketing-related problems and then suggest a relevant decoding scheme for predicting consumers’ choices (e.g., willing to buy or not a specific product). Since the decision-making process involves the concurrent interaction of various cognitive processes and consequently of distinct brain rhythms, the proposed decoder takes the form of an ensemble classifier that builds upon a multi-view perspective, with each view dedicated to a specific frequency band. Adopting a standard machine learning procedure, and using a set of trials (training data) in conjunction with the associated behavior labels (“buy”/ “not buy”), we train a battery of classifiers accordingly. Each classifier is designed to operate in the space recovered from the inter-trial distances of SCMs and to cast a rhythm-depended decision that is eventually combined with the predictions of the rest ones. The demonstration and evaluation of the proposed approach are performed in 2 neuromarketing-related datasets of different nature. The first is employed to showcase the potential of the suggested descriptor, while the second to showcase the decoder’s superiority against popular alternatives in the field.

In this work, we propose a novel framework to recognize the cognitive and affective processes of the brain during neuromarketing-based stimuli using EEG signals. The most crucial component of our approach is the proposed classification algorithm that is based on a sparse representation classification scheme. The basic assumption of our approach is that EEG features from a cognitive or affective process lie on a linear subspace. Hence, a test brain signal can be represented as a linear (or weighted) combination of brain signals from all classes in the training set. The class membership of the brain signals is determined by adopting the Sparse Bayesian Framework with graph-based priors over the weights of linear combination. Furthermore, the classification rule is constructed by using the residuals of linear combination. The experiments on a publicly available neuromarketing EEG dataset demonstrate the usefulness of our approach. For the two classification tasks offered by the employed dataset, namely affective state recognition and cognitive state recognition, the proposed classification scheme manages to achieve a higher classification accuracy compared to the baseline and state-of-the art methods (more than 8% improvement in classification accuracy).

… In the Odor-Induced Emotion Filtering Experiment (OIEFE), 44 participants rated 10 odors for pleasantness, and five odors with high consensus were chosen. EEG data from 9 …

In this study, we propose a hybrid decoding scheme for classifying consumer intent in a binary decision-making scenario (“Buy” vs. “NoBuy”), using simultaneous electroencephalography (EEG) and eye-tracking data. The proposed framework integrates graph signal processing-based features derived from EEG functional connectivity with descriptive statistics from eye movement patterns. Given the imbalanced nature of the targeted classification task, the performance of the proposed hybrid scheme is being assessed at the individual subject level via the employment of Cohen’s kappa and F1-score metrics, both of which are well-suited for handling class imbalance by accounting for agreement beyond chance and balancing precision and recall, respectively. The reported results showcase the superiority of the proposed hybrid decoding scheme, as the averaged scores for both Cohen’s kappa and F1-score are exceeding (with statistical significance at 0.05) the presented competing approaches by 0.08–0.30 and 0.06–0.23 respectively. Additionally, our connectivity analysis confirmed two key findings: (i) strong couplings were consistently observed between electrodes spanning distinct brain regions, such as the prefrontal and occipital cortices, in addition to the commonly reported frontal dipoles; and (ii) the most salient functional connections varied across individuals, with only a limited subset shared among subjects. These results highlight the potential of multimodal decoding approaches and subject-specific connectivity patterns in advancing the classification of consumer decision behavior.

Furaneol and sotolone are caramel aroma isomers with distinct sensory qualities, yet how their differential perception emerges from molecular to neural levels remains unclear. This study integrates sensory analysis, molecular dynamics, and EEG to address this gap. Sensory evaluation showed furaneol was more pleasant and had a tenfold lower detection threshold than sotolone. Simulations revealed furaneol binds stably to OR5M3 via TYR-257, while sotolone interacts selectively with OR8D1 via HIS-159/ASN-206. EEG identified that furaneol uniquely enhanced frontal θ power (4-8 Hz), associated with cognitive engagement, whereas both odorants increased α/β power. These results demonstrate that perceptual differences originate from receptor-specific binding and are cortically encoded in distinct oscillatory patterns. This multi-level approach provides a mechanistic framework linking molecular interactions to perception, supporting the rational design of flavors.

Compared with vision and audition, gustation yields weaker and more confound-sensitive scalp (EEG) responses. Here we review gustatory EEG through an observability framework comprising detectability, reproducibility and interpretability. Low-frequency, especially delta-band, activity currently provides the strongest evidence for taste-identity detectability, whereas event-related potentials (ERPs) evidence is stronger for intensity. Spatial and network features remain promising but methodologically constrained. We outline best-practice workflows for cumulative, interpretable gustatory EEG evidence.

… Particularly, electroencephalography (EEG) technology can objectively evaluate flavor … (OAV) analysis, and integrate EEG to decipher the neuroelectrophysiological differences elicited …