Research On Basic Taste Sensation Recognition Based On Physiological Signals

Taste Sensation,which regulates the eating process and body metabolism,is one of human’s essential sensations.Recognizing taste sensation could be applied in many fields.Traditional sensors aim at recognizing the tastants,however,the taste sensation is affected by not only tastants but also other factors.Physiological signals,which describe physiological activities,can reflect more close-to-fact taste sensations by comprehensively reflecting more factors influencing taste sensations.However,there is little research on recognizing taste sensations based on physiological signals.For the first time,an approach to recognizing basic taste sensations based on physiological signals,especially surface electromyography(s EMG),is proposed in this research.The feasibilities of qualitatively distinguishing different basic taste sensations and estimating the strength of taste sensations of the sour,the bitter,and the salty are verified.The approach includes four steps,data acquisition,preprocessing,feature extraction,and pattern recognition.An experimental protocol and an electrode distribution involving muscles such as masseter,depressor anguli oris,levator labii superioris,risorius,and procerus were designed and optimized.After preprocessing of sample augments,baseline correction,power-line frequency notch,and unqualified sample elimination,features were extracted for pattern recognition methods such as Random Forest and Support Vector Regression.The accuracy of Random Forrest on the 6-labels(5 types of basic tastes and no taste)classification task reached 75.41%,and the coefficients of determinations of the Support Vector Regression on the basic taste of the sour,the bitter,and the salty were 0.7277,0.7450,and 0.7642.Based on s EMG,the feasibility of qualitatively distinguishing different basic taste sensations and the feasibility of estimating the taste sensations strength of the sour,the bitter,and the salty were verified.Moreover,a cross-user model of distinguishing different basic tastes was constructed and optimized with features mapping based on Domain Regularized Component Analysis and samples augment based on Conformal Prediction.At last,the strengths of basic taste sensations were estimated based on the fusion of electroencephalogram(EEG)and s EMG,which verified the feasibility of estimating the taste sensations strength of the sour,the bitter,and the salty based on EEG,and the conclusions that s EMG outperformed EEG,and fusing with EEG patterns could improve the performance of s EMG patterns on the task of estimating the basic taste sensations strength.