Design And Implementation Of Emotion Recognition Software System Based On Physiological Signals

In today's society,with the rapid development of the theory of affective computing,the identification of human emotional state by computer has become a research hotspot.Compared with sentiment analysis based on user's voice,intonation,posture,movement and facial expression data,sentiment recognition technology based on multiple physiological signals can obtain participants' inner expression more truly.However,for the software system in the field of emotion recognition based on physiological signals,there are still some deficiencies.By observing and studying the supporting software of the mainstream physiological signal acquisition device,it is found that:(1)The software system's functions are limited to real-time signal acquisition,display,storage and preprocessing,and it lacks the function of using emotion-inducing materials(especially immersive virtual reality scene materials)to design the presentation mode and sequence of stimuli.(2)lack of emotion recognition data processing function of interfaces containing external data sets;(3)Single function separation,can not use the same software to achieve emotion induction,signal acquisition and emotion recognition functions,the lack of a unified operation interface.Aiming at the above problems,this paper studies the supporting software based on the existing multi-mode physiological signal acquisition device used for emotion recognition in the Mobile Computing Center of University of Electronic Science and Technology of China.The main work contents are as follows:(1)The emotion recognition software system based on physiological signals is designed and implemented.Through the analysis of the physiological signal acquisition device of the current mainstream of form a complete set of software defects and the center of system requirements of the project,in the realization of eeg,pulse wave,skin and skin temperature,physiological signal acquisition,real-time display and storage function,on the basis of design and integration at the same time the following functions:(1)pictures,audio and video,and VR scene material emotion stimulating presentation and sequence function;(2)Offline sentiment recognition functions including external data set interface,including signal preprocessing,feature extraction,feature dimensionality reduction,feature fusion,feature standardization and feature classification;(3)Deploy an online emotion recognition system.(2)An emotional induction experiment paradigm with VR scenes as the material of emotional induction was designed.The subjects' selection,experimental equipment,experimental process and other contents were described in detail.According to the experimental paradigm,the actual test was conducted with software,and the feasibility was verified successfully.(3)Use DEAP data set to conduct practical tests on offline and online emotion recognition functions.(1)The function of offline sentiment recognition was tested,and the effectiveness of offline data processing function of sentiment recognition was verified successfully.(2)Online emotion recognition function test is based on the deployment of the trained model,10 subjects using 8 VR scenes for emotion recognition,the actual test of the valence dicherization accuracy(63.75%)is 6.68% lower than the test set of70.43%.The accuracy of arousal dichotomy(58.75%)was 5.97% lower than that of 64.72%in the test set.Based on software engineering and emotion recognition knowledge based on physiological signals,this paper designs and implements emotion recognition software based on physiological signals,and carries out function and performance tests.The results show that the software has good stability and meets the design requirements.