Evaluation And Detection Of Visual Induced Motion Sickness Based On EEG

In recent years,virtual reality(VR)technology has been widely used in various fields,and its products have gradually integrated into people’s life.With a head-mounted display(HMD)or a large projector screen,virtual reality technology can generate a virtual environment to provide users with abundant three-dimensional information and give people strong three-dimensional feeling and immersion sense.However,people often experience dizziness,nausea,and vomiting when using virtual reality products.We call these symptoms visually induced motion sickness(VIMS)or virtual reality sickness(virtual reality motion sickness,VRMS).These symptoms seriously affect the user experience of virtual reality products,which greatly restricts the further development of virtual reality technology.Therefore,it is of great significance to carry out the research on VIMS.The purpose of this dissertation is to analyze the relationship between electroencephalogram(EEG)and visually induced motion sickness level(VIMSL),and to find the EEG marker that can be used to evaluate VIMS;on this basis,the automatic detection of VIMS symptoms is achieved.This dissertation first designs a VR-based vehicle-driving simulator(VDS)experiment,which can induce subjects to produce VIMS symptoms;then uses the Muse device to collect the subjects’ brain electrical signals(EEG)data.Subjective VIMS level(VIMSL)data that reported by the subjects themselves orally were collected at the same time;finally,the VIMS was analyzed using a combination of subjective method and objective method to find the VIMS evaluation indicators based on the Kolmogorov complexity(KC),and a VIMS detection algorithm based on machine learning was proposed.The main research results of this dissertation include:(1)VDS experiments based on VR are designed.The experiment included a baseline stage,a driving stage,and a resting stage.In these three stages,the wearable wireless device Muse was used to collect the EEG data,and subjective VIMSL data in the experimental process was obtained through the subjective oral report.Experiments show that this experimental scheme can invoke severe VIMS symptoms for most subjects.(2)In the aspect of VIMS evaluation,this dissertation finds the VIMS evaluation marker based on KC complexity by analyzing the correlation between EEG marker and subjective VIMSL and comparing whether there is a significant difference between EEG marker before and after motion sickness.The experiment shows that: 1)the KC complexity of the 4 channels(TP9,FP1,FP2 and TP10)EEG correlation with VIMSL’s Spearman rank correlation analysis shows that there is a certain correlation between them.2)The KC complexity of the three channels(TP9,FP1,and FP2)was significantly different before and after VIMS,and the KC complexity was significantly decreased after VIMS occurred,indicating that the KC complexity of the three channels can be used as an evaluation marker.3)After VIMS occurred,70% of subjects had a significant decrease in the KC complexity of the TP10 channel,while the KC complexity of the 30% subjects did not descend;this may be caused by the individual difference of the subjects.In the aspect of VIMS detection,this dissertation proposes a VIMS detection algorithm.Firstly,the algorithm preprocessed the EEG data.Then it extracted the features based on the preprocessed data and selected the feature subset with high correlation with VIMSL by feature selection.Finally,the state of the VIMS was classified by constructing the classifier.The experiment shows that: 1)The feature extraction and selection method of this dissertation can effectively extract the characteristics of VIMS reflected in EEG data;2)Compared with support vector machine(SVM)and BP neural network,Bagging-based random forests(RF)classifiers can achieve higher detection accuracy(76.5%)result.The research results show that the KC complexity can be used as a marker to evaluate the VIMS;the VIMS detection algorithm proposed in this dissertation can achieve high detection accuracy in the case of only four channels of EEG data,which is suitable for the construction of VIMS detection system under the condition of less channel wearable wireless devices.The research results in this dissertation help to further understand VIMS and enrich the research on VIMS,so as to provide references for the improvement of virtual reality technology.