Modeling And Analysis Of Human Emotions Under The Condition Of Human-robot Interaction Based On Virtual Reality

With the acceleration of the aging of the whole population structure in the world,the old-age care problem will become more and more intense and become one of the problems that can not be ignored by the international community.In recent years,with the rapid development of robot technology,some intelligent service robots have been equipped with the ability to care for the elderly in daily life.In the process of completing the task,although the robot can basically complete the care task,it usually does not consider the subjective feelings and psychological needs of the objects served,and cannot achieve the best service effect.This paper aims to establish a emotional recognition model,the robot according to the emotional recognition model to perform a task in the process,able to recognize human emotions and can according to individual psychological status and subjective feeling to conform to the will of people to adjust,avoid negative influence on user's mood,improve compatibility between robot perception and user expectations,will help the robot more active,efficient,safe,comfortable service for users.In order to improve the completely objective information quality evaluation mode in the traditional emotion modeling and improve the subjective effect of the traditional cognitive evaluation,a dynamic emotion evaluation system was constructed based on the above emotion recognition model,which centered on emotion,emotional subjective evaluation and emotional physiological response.Based on Russell's two-dimensional emotional model and self-assessment model SAM,the system establishes a set of training methods to screen subjects who can correctly digitally express their emotions based on external stimuli.Then to understand man-robot interaction of a single variable stimulate emotional impact to human body,combining the wheelchair robot and virtual reality to build a kind of can assess the independent factors of immersive interactive experimental platform,the experiment for a variety of emotional physiological data(ECG,EMG,skin conductance,and frontal lobe brain blood oxygen content)and its corresponding SAM model evaluation level,thus making arousal data calibration database.IAPS images with the same or highly similar valency and arousal with the subjective evaluation obtained in the emotional arousal experiment were selected based on the subjects' SAM evaluation.The emotional induction was carried out again,and the results were verified with the above experiments,so as to build a comparative relationship between the two experimental data.Finally,subjective evaluation was used as the intermediate criterion between expressive representation and emotion label prediction,and the mapping relationship between multimodal physiological signals of autonomic nervous system and central nervous system and correct emotion expression was constructed by using CNN deep learning method.The overall recognition rate of CNN emotion classification model reached 82.5%.At the same time,the performance of the model was analyzed,and the sample of emotional arousal database was predicted.The results show that the model after training can accurately label the emotion represented by the physiological signal sample,and realize the engineering application of emotion recognition based on multiple physiological signals.It also forms a new emotional evaluation method combining subjective and objective.