Research On User's Emotion Recognition Model In Virtual Scene

Human emotion is complex and diverse,mainly through facial expressions,gestures,voice and other emotional representations to convey emotion.In today's artificial intelligence development and virtual application scenarios,the research of"affective computing" that uses computers to identify,understand and express emotions,especially the key problem of emotional computing,"emotional recognition" is becoming the basis of building a harmonious human-computer emotional interaction environment,and it is also the analysis of human external emotional performance behavior,human physiological signals and other internal emotional information to infer human emotions,the hot research topic of state.This paper is based on the research that physiological signals are different from other external emotional representations and controlled by the activities of the ANS,which can more objectively reflect the real emotional state of users.It designs and constructs the collection environment and scene of physiological signals such as ECG,EDA,EMG,PPG,etc.A physiological signal acquisition scheme in virtual scene is built,and the positive and negative emotion states of users are investigated by using PANAS emotion scale to evaluate the effect of scene induction.According to the needs of noise and feature recognition of four kinds of physiological signals collected by the subjects:ECG,EDA,EMG,PPG,the time-domain feature,frequency-domain feature and non-linear feature are used to extract the signal features,the MMAS algorithm is used to select the features,and the simple Bayes,decision tree,k-nearest neighbor and support direction are used to select the features Four classical classifiers are used to identify the emotional states of subjects,such as pleasure,surprise,fear,disgust,and so on.SVM is used as classifier to analyze the difference between statistical features and non-linear features in the effect of emotion recognition.In addition,the method of subjective and objective combination is used to identify the user's emotion,and a user's emotion model suitable for virtual scene and a good accuracy of emotion recognition are established,which has a high guiding role and application value for the future emotion recognition research.