Realization Of Individual Behavior Online Evaluation And Prediction System

Nowadays,Brain Computer Interface(BCI),as a multidisciplinary interdisciplinary technology,is gradually attracting the attention of scholars in various fields.Due to the rapid development of brain science research,many researches on brain science have obtained great achievements.The most direct way to apply the achievements of brain science to reality is to develop BCI technology.According to the main work content during my master's degree,I propose a brain-computer interface system that can monitor the behavioral state of individuals in real time.This system aims to achieve online assessment and prediction of individual physiological and psychological behavior through BCI technology.The system development mainly includes:1?Online recognition of different finger action tasks.The system uses signal acquisition equipment to monitor individual arm Electromyography(EMG)signals in real time,and extracts Common Spatial Pattern(CSP)features after preprocessing the EMG signals.This feature is used to train and test Linear Discriminant Analysis(LDA)to complete the recognition of different finger activity tasks.Based on different application scenarios,the system has developed two test modes: synchronous test mode and real-time test mode.In the synchronous test mode,the system is used to identify the user's actions when there is a task prompt.In real-time test mode,the system is used to identify the spontaneous actions of users in real time.The main difference between the two modes is that the task state of EMG signal is monitored in the real-time test mode.By judging the task state,the system can automatically collect the EMG signal after the user's finger movements for recognition,so as to realize the real-time monitoring of finger movements.We recruited 11 subjects to test the sub-system.The average accuracy of 6 types of gesture recognition for all subjects in the synchronous test mode was 90.682%;the average accuracy in the real-time test mode was 94.669%.Among them,the average time to complete a classification task in the synchronous test mode is 3s,while the real-time test mode is only 0.6s,which indicates that the system has high classification accuracy and real-time characteristics.2?Real-time monitoring of individual emotions.The system uses video materials to induce emotions of the subjects,and monitors individual Electroencephalogram(EEG)signals through signal acquisition equipment.Coherence analysis(COH)method was used to construct EEG networks between different lead signals,then power spectrum and network features were extracted,and feature fusion was used to select the optimal feature set.The LDA classifier performed training and real-time emotion recognition based on this feature set,and finally output and display the decoded emotion results.So as to realize the purpose of using the system to monitor the individual emotional state in real time.In order to evaluate the performance of this sub-system,we recruited 14 subjects for emotion testing experiment.The average online decoding accuracy of the system for the two types of emotions was 80.25%.The decoding accuracy rate is high,and the individual differences between different subjects can be effectively eliminated,and it has high stability.