Working Memory Cognitive Training And Assessment System With EEG Signal Analysis Based On BCI-VR

Working memory ability is closely related to various activities of daily life.Training and evaluation of working memory ability is a valuable research direction.The traditional working memory training is simple,which is easy to be disturbed by environment,and the evaluation of working memory cognitive has certain subjectivity.Training working memory ability based on virtual reality and brain-computer interface digital games has become a research hotspot in the field of cognitive science at home and abroad.In this paper,a virtual working memory training system is designed based on the theoretical model of working memory and proposed Normalized and Permutation Conditional Mutual Information,to realize the effective evaluation of EEG signal.First of all,this paper builds a working memory training evaluation system integrating virtual reality and brain-computer interface,which includes scale link,training link and test link.Based on the theoretical model of working memory and Philips Vision short-term memory task,VR training and testing games for working memory were designed to conduct cognitive training for 33 days.DST and CBTT tasks were introduced to quantitatively analyze the working memory ability before and after training.EEG signals were collected by brain-computer interface technology.EEG features were extracted by sequencing conditional mutual information method,and the feature extraction results were visually described by brain network technology to analyze the changes of working memory.The random forest method is used to classify the EEG features and analyze the training effect of the system comprehensively.Secondly,in order to study the correlation relationship and changes of EEG channels in task-driven working memory,proposed a EEG feature extraction method,which named Normalized and Permutation Conditional Mutual Information.This method takes into account the influence of other surrounding channels on the combined channels,and limits the infinite growth of results caused by excessive number of cut samples.On the basis of ensuring the accuracy of feature results,it adds feature diversity to accurately describe the EEG activity under the working memory task.Finally,14 students were recruited to participate in the experiment.The training effect of the virtual library working memory cognitive training was evaluated comprehensively from the three aspects of the participants’ scale scores,behavioral data and EEG,and the effectiveness of the system was verified.The CNN method was used to classify the EEG features extracted by NMPCMI and compared with the traditional mutual information algorithm to verify the validity of the model.