Research And Implementation Of Emotion Recognition Fusion Technology Based On Physiological Signals

With the development of computer technology,people spend more time communicating with computers than with people.Therefore,analyzing and recognizing people’s emotional state through computer has become a new field and focus of research.At present,researches on facial expressions,speech emotions,human postures,and text emotion recognition have been carried out at home and abroad,and preliminary results have been obtained.However,there are some drawbacks in these emotional expressions.They can be hidden or changed through artificial subjective thoughts,so that they can not fully express the real emotional state of human beings.Emotional recognition based on human physiological signals has objectivity and can not be subjectively manipulated,but it has better recognition effect on emotions with high arousal.Therefore,this thesis proposes an emotional recognition method based on subjective and objective fusion,which makes the subjective and objective emotional advantages complement each other in the recognition process to improve the emotion recognition effect.In the objective aspect,two kinds of physiological signals,Galvanic Skin Response(GSR)and Heart Rate,are selected and analyzed to extract the statistical characteristic values of the two kinds of signals under four basic emotional states of happiness,sadness,anger and fear,such as mean,standard deviation,median,maximum value,minimum value,first-order difference,second-order difference,etc.Then use the Support Vector Machine(SVM)-based classification model to identify and train the feature set,and obtain the decision information of four basic emotional states.In the subjective aspect,subjective emotions in four emotional states are used as data sources,quantify the subjective emotion by labeling the PAD emotion scale,and get the three-dimensional emotional values of PAD,and its distribution in the three-dimensional emotional space of PAD is obtained.We can get the decision information of emotion by measuring and comparing the distance from the standard position of the emotional states.Because different classification methods are used,and the subjective and objective data sources have different recognition effects for the four emotions,this thesis proposes a method of combining subjective and objective decision information,and uses Particle Swarm Optimization Algorithm(PSO)to optimize and assign the weights of the four emotional states of each classification method,and construct a subjective and objective fusion emotion recognition model.Experiments show that the effect of fusion of subjective and objective emotion recognition is better than subjective or objective emotion recognition.Finally,this thesis applies the established subjective and objective fusion model to the user experience emotion recognition,and constructs an emotion system based on user experience,which integrates the objective physiological signals and subjective emotions generated by the user during the viewing to identify the emotional state of the user experience and evaluate the user experience.