Study On EEG Mechanism And Classification Under Digit Working Memory Task

Working memory is very important in brain research,and it is one of the key components of human thinking and behavior.Working memory plays an important role in learning and education.With the deepening of informatization,in order to save time,learners need to learn knowledge by increasing the playback speed.Double-speed playback will speed up the presentation of information,thereby challenging people’s working memory.Therefore,this paper aims to explore the impact of playback speed on working memory.It is expected that it can be applied to teaching situations and provide a theoretical basis for teachers to design playback speeds suitable for students’ learning.Create a more efficient learning environment.First of all,due to the lack of EEG databases for different visual presentation speeds,in order to induce working memory states and collect data,a new experimental paradigm was designed to induce working memory by memorizing numbers.In this experiment,nine numbers were played at different visual presentation speeds,and the EEG data of 18 subjects were collected to establish a working memory EEG database at fast and slow presentation speeds,and conduct behavioral analysis on the accuracy of working memory in fast and slow states,found that the average correct rate of the subjects in the slow state was much higher than the average correct rate of the subjects in the fast state,and the difference was extremely significant(P=1.021e-05,P<0.01),and then the brain in the fast and slow state Frequency-domain and time-frequency domain analysis of electrical signals were performed to study its internal mechanism.It was found that the power spectrum of the frontal lobe was suppressed at the fast visual presentation speed in the theta frequency band,and the power spectrum of the inferior occipital lobe was suppressed in the alpha frequency band.Secondly,due to the mutual influence and causal relationship between different brain regions,in order to further study the operating mechanism of working memory,we constructed a causal brain network of characteristic frequency bands according to the Granger causality between channels,and proposed that from the brain network From the perspective of working memory in the fast and slow state,extract the network characteristics(ingress,clustering coefficient,network efficiency)of theta,alpha,and beta frequency bands in different working memory states to mine the characteristics of working memory in the fast and slow state.The results show that in In the state of fast visual presentation,the ingress and exit degree of the brain network increases,the node clustering coefficient and network efficiency are further strengthened,and the nodes with significant differences are mainly distributed in the frontal,parietal and occipital lobes.This study shows that as the speed of visual presentation increases,The visual processing was further activated,the subjects’ working memory awareness activities gradually increased,and the leading role of the left hemisphere of the brain in cognitive activities such as language and reasoning was also continuously strengthened.Finally,this paper studies the working memory in the fast and slow state from the perspective of entropy.Since the brain itself is a chaotic system,and entropy can quantify the complexity of the network,the network entropy theory is introduced,and the network based on the weighted and directed network Features(in-degree,out-degree,clustering coefficient)calculate network entropy.Calculate the network entropy of the global and local brain regions respectively,and analyze the global network entropy of theta,alpha and beta frequency bands and the local network entropy of each brain region under different working memory states.This study shows that as the speed of visual presentation increases,the entropy value is increasing,and the entropy value in the beta frequency band increases significantly.And use the support vector machine to classify the fast and slow visual states through the network characteristics and network entropy of each frequency band,and the classification accuracy can reach up to 90.96%.