Research And Implementation Of Emotion Recognition And Prediction System Based On Multiple Physiological Signals

Nowadays,the time of communication between people and computers far exceeds the exchange between people and the rise of the theory of emotional computing.The recognition of emotional states through computers has become the focus of research.At present,a lot of emotion recognition is based on the user's voice intonation,facial expressions,body movements and other information as a data source for analysis,to identify the user's emotional state,but these expressions have some drawbacks,such as can be controlled by human subjective thinking,It is not enough to fully express people's true emotional state.Moreover,many of the recognitions of emotional states are aimed at real-time emotion recognition.They do not predict the emotional changes in the short term and cannot take measures to predict the emotional state in advance.To prevent the occurrence of some hazards.Therefore,for the source of emotion recognition dataset and emotional prediction problem,this thesis proposes two methods of emotion recognition and prediction based on multiple physiological signals.The emotion recognition method is to analyze the two types of signals of skin electrical and pulse signals and extract the two types of signals.,then use the BP neural network algorithm-based emotion recognition model to identify and train the feature set,to obtain four basic emotional states,and to solve the subjective control problem of the data set.In order to solve the convergence speed and recognition rate of the emotion recognition method,a GA-AdaDelta-BP recognition algorithm based on GA,AdaDelta algorithm and BP algorithm is proposed.Based on the recognition of emotions by BP algorithm,GA determines the initial of BP algorithm,AdaDelta algorithm does BP algorithm's adaptive learning rate,effectively improving the convergence speed and recognition rate.The emotion prediction method is based on the nonlinear characteristics of the pulse signal,uses a prediction algorithm based on BP neural network to predict the pulse signal sequence,and then uses the improved emotion recognition method to analyze and identify the emotional state,that is,the future emotional state.In order to solve the accuracy problem of the emotion prediction method,GA,AdaDelta algorithm and BP prediction algorithm are combined together,and a GA-AdaDelta-BP prediction algorithm is proposed,which effectively improves the accuracy of the prediction algorithm.Finally,this thesis applies the proposed two algorithms to physiological signal processing and constructs an emotion recognition and prediction system.After experimental analysis,the recognition rate and accuracy of the optimized algorithm have been significantly improved.