The Effect And Evaluation Method Of Virtual Reality On Target Recognition With EEG

EEG(Electroencephalogram)is a kind of electric potential generated from the activity of human brain nerve cells.The 2D target recognition based on EEG basically adopts the rapid serial visual presentation paradigm(RSVP)or Oddball paradigm(Oddball).The specific EEG component P300 is used to achieve rapid recognition of the target image.The peak and latency of P300 can objectively reflect the subject's cognitive ability,such as acceptance,processing and response to stimuli.Therefore,the recognition of P300 component in EEG is the core and the key issue of 2D target recognition based on EEG.This study will start with the characters of the P300 component and on this basis,we further improve the efficiency of P300 component recognition from the perspective of signal processing,thereby enhancing the effect of target recognition based on EEGAt present,the current situation of using VR for quantitative research on target recognition with EEG is still missing.In this study,VR is used in the research work of target recognition with EEG.Identifying targets through EEG and VR devices has important research significance:on the one hand,it can be experimentally explored whether it is easier to induce P300 signals under VR;on the other hand,for the subjects,whether the performance under 3D and VR is consistent and whether the effects on target recognition with EEG are the same in all modes is also worth studying.The research work done in this article includes(1)Use the classic induction paradigm of target recognition Oddball to design a standard experiment,collecting the EEG signals of the subjects under the 2D target recognition experiment as the standard data set to train the LSTM classifier,then use the classifier to predict the target recognition level classification label in 3D mode and VR mode.(2)Using the 3D mode and VR mode of the same task as a comparison,design a control experiment,exploring the differences between the individual's target recognition level with EEG in 3D mode and VR mode,through the experimental results to verify whether the beneficial hypothesis theory proposed by the study is correct.(3)In terms of data processing,a method of multi-channel EEG artifacts filtering based on ICEEMDAN-ICA is proposed for P300 signal collection and good results are obtained compared with similar algorithms.In the feature extraction stage,we use wavelet transform,time-domain energy entropy and FastICA methods to carry out a full range of feature extraction of P300 in time domain,frequency domain and spatial domain,finally build an SVM classifier based on machine learning and a two-way LSTM classifier based on deep learning for classification,where the average classification accuracy of the SVM classifier reached 90%and the average classification accuracy of the two-way LSTM classifier reached 95.8%.(4)Combining the performance data and EEG data to analyze the target recognition level of the subject in 3D mode and VR mode,finally verify the VR mode is significant for the target recognition level of the subject through statistical analysis methods such as paired T test influences.By processing and analyzing the EEG signals of the subjects,the following conclusions are obtained:Compared with the 3D mode,the subjects have better performance in the VR mode and can induce a more obvious P300 signal,besides the gender difference has no significant effect on target recognition level.The above conclusions can provide some theoretical support for applying VR scenes to target recognition and other fields.