Research On Personality Recognition Based On ECG Signal In VR Scene

In recent years,personality recognition based on physiological signals has attracted the attention of researchers in psychology and computer science.Previous researches on personality recognition based on physiological signals mainly focus on the nonimmersive 2D video emotional stimulation scene,and there is no research on the immersive Virtual Reality(VR)video emotional stimulation scene.Due to the high immersion and other advantages of VR technology,it has made more and more practical applications in many fields.In order to facilitate the use of wearable devices to detect personality traits,this paper designs a way to analyze the personality traits of Electrocardiogram(ECG)signals collected in VR scenes,and uses machine learning algorithms to construct big five personality recognition models in VR and 2D scenes to realize the classification and recognition tasks in VR and 2D video scenes.Moreover,the statistical analysis results and personality recognition results of VR and 2D video scenes are compared,which reveals the advantages of VR video as emotional stimulation material.This paper mainly studies the design of emotion induction experiment,the acquisition and pretreatment of ECG signal,feature extraction and the construction of personality recognition and classification model.The main research contents and results are as follows:(1)An emotion-personality physiology database EP-SWU based on ECG signals is constructed.In order to construct the recognition model,the primary task is to solve the current lack of a database of personality physiology in VR scenarios.This paper proposes a method to obtain ECG data by using VR and 2D video as emotional stimulus materials,and constructs a physiological database EP-SWU.ECG signal data were collected from103 subjects under seven emotional stimulus videos in VR and 2D groups.The ECG data were divided into two categories by the scores of the Big Five Personality Inventory(NEO-FFI)filled by the subjects,and the ECG data were labeled with different personality labels.This database provides data support for subsequent personality recognition modeling and related research.(2)A machine learning model for personality recognition is constructed.This paper proposes a method to recognize the personality of moviegoers in immersive VR scenes and non-immersive 2D scenes based on ECG signals.Firstly,the ECG in the EP-SWU database was preprocessed to obtain the RR interval sequence.Then,time domain,frequency domain and nonlinear features were extracted from RR interval sequence.Finally,the normalized feature data were input into three machine learning algorithm models respectively,and ten-fold cross validation was used for training and classification,and the sequence Backward Selection algorithm was used for feature selection.The experimental results showed that for the VR group,the binary classification results(F1score)of the classification model in the dimensions of neuroticism,extraversion,openness,agreeableness and conscientiousness were 0.79,0.80,0.81,0.73 and 0.81,respectively.For the 2D group,similarly,the highest results for personality classification were 0.75,0.76,075,0.81,and 0.78,respectively,all significantly higher than random guessing.This shows that personality recognition based on ECG signals and machine learning algorithms is feasible in immersive VR scenes and non-immersive 2D scenes,and can achieve high recognition effects.(3)The advantages of identifying personality traits in VR are compared and verified.The statistical analysis results and classification model recognition effects of the VR group and the 2D group were compared,which revealed the advantages of VR in the field of personality computing.At the level of statistical analysis,the correlation analysis of the personality scale of the VR group and 2D group was first carried out respectively.The results showed that there were similar rules in the personality scores of the two groups,which was consistent with the results of previous studies.Then the emotional arousal scale(SAM)of the two groups was analyzed,and the results verified the effectiveness of emotional arousal.Then the correlation analysis of the NEO-FFI scale and SAM scale of the two groups was carried out,and the results showed that the VR group and the 2D group had different emotion-personality correlation laws.Finally,the correlation analysis of ECG features and NEO-FFI scale between the two groups was carried out.The results show that the VR group has more significant correlation features than the 2D group,which indicates that the VR group reflects more physiological correlation between emotion and personality.In terms of the recognition effect of the classification model,it was found that the recognition results of the VR group in the personality dimensions of neuroticism,extraversion,openness and conscientiousness were higher than those of the2 D group,which indicated that the use of VR as the presentation method of emotional stimuli could better reveal personality differences.The results of this paper show that the feasibility of using ECG signals to identify personality in immersive VR scenes and non-immersive 2D scenes.The significance is that it can be used to improve the user experience of VR and 2D viewing scenes,realize the identification task of user personality traits,and provide corresponding personalized services and guidance,which has broad application prospects.At the same time,this paper compared the differences between VR and 2D in the presentation of emotional stimuli from two levels of statistical analysis and classifier recognition effect,and verified the advantages of VR in the field of personality computing,which provides a useful reference value for future research on personality and emotion based on VR.