Research On Recognition Of Movement-related Channels Based On Brain Function Network

Exercise-induced fatigue can cause different degrees of damage to healthy people and patients with motor dysfunction,therefore,how to effectively detect exercise-induced fatigue has attracted a lot of research.Exploring the relationship between brain regions based on EEG signals to detect exercise-induced fatigue is a hotspot in current research.However,there are different functional regions and many channels in the brain,selecting channels more related to movement from a large number of channels not only reduces computational complexity,but also improves fatigue detection accuracy.Previous studies have identified movement-related channels based on single-channel amplitude changes,these studies were limited to a single channel,ignoring the interaction between channels.However,different functional regions of the brain interact with each other,forming a complex system to accomplish the corresponding tasks.Compared to single-channel features,brain networks constructed based on interactions between channels have proven to be more structurally stable.Based on this,this thesis will identify movement-related channels based on brain networks.The main work and results include:(1)Two types of brain function networks are constructed based on phase synchronization(PLV),and then global network topologies and local network topologies are extracted,combined with Borg value(score of fatigue degree)to carry out the identification research of movement-related channels.1)The global network topologies significance test results show that there are large differences in the global topologies of the different age groups,so the adults and children in the subjects are separated to discuss the identification of the movement-related channels.2)This thesis uses the method of Pearson significant correlation for channel selection,that is,the channel whose local network topologies are significantly correlated with Borg value is regarded as movement-related channel.3)Channel selection based on the Relief algorithm and comparison with channel selection results based on Pearson significant correlation.(2)The selection results of the above movement-related channels are verified under two strategies,namely,the verification strategy based on phase synchronization,the effect of fatigue detection is evaluated by observing the situation of the change of PLV of channel pairs with Borg value,and the classification-based verification strategy,the effect of fatigue detection is evaluated by comparing the classification accuracy of local network topologies.In the above two strategies,the verification is divided into two parts.1)The internal verification of channel selection based on Pearson significant correlation,the movement-related channels are used as the experimental group,and the same number of non-movement-related channels are used as the control group,and the fatigue detection effects of the two groups are compared.2)The verification of the comparison of channel selection based on Pearson significant correlation and the Relief algorithm,that is to compare the fatigue detection effect of the two selected movement-related channels.The experimental results show that the fatigue detection effect of the movement-related channels selected based on Pearson significant correlation is better in the two verification strategies,the validity and reliability of the method of channel selection based on Pearson significant correlation and movement-related channels identified by this method are verified,the results of movement-related channel selection provide a reference for researchers studying the relationship between exercise-induced fatigue and brain regions,and have important significance for the detection and evaluation of exercise-induced fatigue based on EEG signals.