Emotional Physiological Perception Modeling And Channel Optimization Analysis Based On Virtual Reality Headset

In recent years,with the development of virtual reality technology,effective emotion perception in virtual reality environment has become a new research hotspot.Compared with traditional emotional experimental media,VR helmets can provide emotionally induced scenes with better immersion and interactive experience.Scientists in the fields of bioengineering and psychology use VR headsets as media playback devices for emotional physiological computing research;in addition,VR headset users’ interactive information is important,emotion perception under VR environment has also attracted great attention from the industry.Based on this,thesis focuses on several issues in emotional physiological perception based on VR headsets.The main research work is as follows:(1)Research on emotional physiological perception channels in VR environment.In view of the cumbersome EEG signal acquisition process and redundant channels in the VR environment,thesis proposes an emotional physiological perception channel scheme for VR headsets—two-channel EEG on the forehead and three-channel peripheral physiological signals.The scheme performs modeling and channel substitution and enhancement analysis on the open source dataset DEAP.The experiment verified that the2-channel EEG(Fp1,Fp2)has a good ability to represent emotions by performing power spectral density,channel correlation analysis of differential entropy and quantitative analysis of channel importance based on the Gini index on the 32-channel EEG of the whole brain.Multi-channel EEG has a substitution effect.At the same time,the program is compared with other emotional physiological perception channel programs in the field to verify the effectiveness of the program in emotional perception,and the enhancement effect of peripheral channels on EEG channels.(2)Development of emotional physiology experimental software platform and preparation of VR emotional physiology dataset.During the preparation of VR emotional physiology experiments,problems such as the lack of effective VR scene library and inconsistent experimental software platforms were found.In thesis,the selection and evaluation experiments of VR scenes were carried out,and 8 VR emotion-induced scenes were screened;an integrated experimental software integrating the functions of experimental paradigm design and signal acquisition was developed.Using the software platform and the hardware platform of the laboratory to carry out experiments,a VR multi-modal emotional physiological dataset VREPD was produced.(3)Effective recognition of physiological emotions.In the process of using a VR helmet,there may be abnormal EEG signals accompanied by head and body movements,as well as the problem of mining deep emotional information from multi-modal physiological signals.Thesis proposes an Ada Boost channel controller + CNN-LSTM network.Through the CNN-LSTM space-time network,the network can effectively mine the deep temporal and spatial emotional information of multimodal signals,and through the Ada Boost channel controller,it can effectively identify abnormal EEG caused by movement and perform channel control.When the EEG is abnormal,the network automatically controls the decision-making channel and only uses peripheral physiological signals for emotion recognition.Using this network to conduct experiments on the DEAP data set and VREPD data set containing EEG bad segments,the binary classification accuracy of valence and arousal reached 0.8510 / 0.8617(DEAP)and0.9051 / 0.9314(VREPD),respectively,proving its Effectiveness of emotion recognition during dynamic use of a VR headset.