Abstract:The taste electroencephalogram (EEG) evoked by the taste stimulation can reflect different brain patterns and be used in applications such as sensory evaluation of food. However, considering the computational cost and efficiency, EEG data with many channels has to face the critical issue of channel selection. This paper proposed a channel selection method called class activation mapping with attention (CAM-Attention). The CAM-Attention method combined a convolutional neural network with channel and spatial attention (CNN-CSA) model with a gradient-weighted class activation mapping (Grad-CAM) model. The CNN-CSA model exploited key features in EEG data by attention mechanism, and the Grad-CAM model effectively realized the visualization of feature regions. Then, channel selection was effectively implemented based on feature regions. Finally, the CAM-Attention method reduced the computational burden of taste EEG recognition and effectively distinguished the four tastes. In short, it has excellent recognition performance and provides effective technical support for taste sensory evaluation.

What problem does this paper attempt to address?

This paper aims to solve the key problems in food taste sensory evaluation, especially how to optimize the taste electroencephalogram (EEG) channel selection through neural network technology to improve the efficiency and accuracy of food sensory evaluation. Specifically, the paper proposes a channel selection method that combines convolutional neural network with channel and spatial attention mechanisms (CNN - CSA) and gradient - weighted class activation mapping (Grad - CAM), called the Class Activation Mapping and Attention (CAM - Attention) method. This method can not only reduce the computational burden of taste EEG recognition, but also effectively distinguish four basic tastes (sour, sweet, bitter, and salty), thus providing effective technical support for food taste sensory evaluation. ### Main contributions of the paper: 1. **Designed the taste EEG experiment**: Collected taste EEG data using different taste stimuli. 2. **Proposed the CNN - CSA model**: This model can fully utilize the key features of taste EEG. 3. **Developed the CAM - Attention method**: Combined the CNN - CSA and Grad - CAM models, effectively selected the key channels of taste EEG, mined data features in the selected channels, and finally achieved effective distinction of the four tastes while reducing the amount of computation. ### Experimental methods: - **Data collection**: Used 21 electrodes to collect taste EEG data from 20 subjects, and the subjects tasted four solutions of sour, sweet, bitter, and salty respectively. - **Pre - processing**: Performed pre - processing operations such as band - pass filtering and down - sampling on the collected EEG data. - **Analysis framework**: Trained the classification task through the CNN - CSA model, then used Grad - CAM to generate the gradient class activation map, and finally selected the key channels according to the activation map. ### Results and discussion: - **Model settings and evaluation methods**: Randomly divided all taste EEG samples into training sets and test sets, and used accuracy and F1 - score as evaluation indicators. - **Performance evaluation**: Compared with traditional channel selection methods such as Lasso, MI - BMInet, and NMI, the CAM - Attention method can still maintain a high classification accuracy and F1 - score while reducing the number of channels. For example, when 12 channels are selected, the accuracy of the CAM - Attention method is 97.55%, the F1 - score is 97.52%, which is only 0.44% and 0.43% lower than the performance in the case of all channels. In conclusion, through proposing the CAM - Attention method, this paper effectively solves the computational efficiency and classification performance problems in taste EEG channel selection, and provides a new technical means for food taste sensory evaluation.

Optimizing food taste sensory evaluation through neural network-based taste electroencephalogram channel selection

Basic Taste Sensation Recognition from EEG Based on Multiscale Convolutional Neural Network with Residual Learning.

Bimodal Fusion Network for Basic Taste Sensation Recognition from Electroencephalography and Electromyography

Basic Taste Intensity Recognition Based on Semg and EEG Signals

Decoding Taste Information in Human Brain: A Temporal and Spatial Reconstruction Data Augmentation Method Coupled with Taste EEG

Feature Extraction and Feature Combination for Basic Taste Sensation Recognition Based on Facial Electromyography

Decoding human taste perception by reconstructing and mining temporal-spatial features of taste-related EEGs

Basic Taste Sensation Recognition from EEG Based on Multi-Scale Convolutional Neural Network with Residual Learning

Qualitative Recognition of Primary Taste Sensation Based on Surface Electromyography

Quantitatively Recognizing Stimuli Intensity of Primary Taste Based on Surface Electromyography.

An in Vivo Bioelectronic Tongue for Sweetness Detection Using Rat Gustatory Perception Based on Brain-Computer Interface

TastePeptides-EEG: an Ensemble Model for Umami Taste Evaluation Based on Electroencephalogram and Machine Learning.

Advancing research on odor-induced sweetness enhancement: A EEG local-global fusion transformer network for sweetness quantification combined with EEG technology

Insights into Brain Perceptions of the Different Taste Qualities and Hedonic Valence of Food Via Scalp Electroencephalogram

An Effective Identification of Flavor Complaint By Adaptive Analysis of Electroencephalogram (EEG) Signal

Recognition of odor and pleasantness based on olfactory EEG combined with functional brain network model

A Comprehensive Framework for Decoding Salty Taste Information from EEG Signals: Distinguishing Brain Reactions to Saltiness of Comparable Intensity

Unsupervised cross-user adaptation in taste sensation recognition based on surface electromyography with conformal prediction and domain regularized component analysis

Electronic Tongue Based on Ion-selective Electrode and MCU

Flavor Identification Based on Olfactory-Taste Synesthesia Model and Hybrid Convolutional Neural Network-Random Forest

Enhancing EEG Signal Classification With a Novel Random Subset Channel Selection Approach: Applications in Taste, Odor, and Motor Imagery Analysis