EEG-Based Taste Perception Classification Using PCA Enhanced Attention-TLSTM Neural Network

EEG-based taste perception detection offers a unique and valuable insight into how taste perception occurs. These methods find applications in both clinical and research environments. These techniques can be utilized for the diagnosis of different taste-related disorders, such as ageusia, hypogeusia, dysgeusia, etc., and to assess how medical treatments impact taste perception. In humans, a loss of taste sensation may indicate the presence of a neurological disorder. Detecting neurological disorders as early as possible is very important. The essence of this paper is to propose a novel deep learning technique for differentiating six distinct taste perceptions, i.e., no taste, sour, sweet, bitter, salty, and umami, using the Electroencephalography (EEG) signals of the human brain. First, the EEG signals of the participants have been captured meticulously, and subsequently, different features have been extracted from the recorded EEG signals. Then the principal component analysis (PCA) technique has been employed to reduce the dimensionality of the extracted features. An attention module-incorporated TLSTM (tanh-long-short-Term-Memory)-based novel classifier has been designed and trained using the optimal EEG features. The designed classifier model attains an impressive average classification accuracy of 97.16%, coupled with precision and recall rates of 97.5% and 97.1%, respectively.