Research On Motion Imaging Pattern Recognition Method Based On Brain Function Network

Motor Imagination(MI)is a new method of rehabilitation,it is as effective as the actual exercise to activate the motor-related cerebral cortex.The brain function network can effectively reflect the activity state of the brain,and it can be used as an auxiliary method for Motor Imagination research.In recent years,a large number of researchers have devoted themselves to the study of pattern recognition methods for motor imagery EEG signals,this technology is one of the important technical indicators in the fields of rehabilitation robots and brain-computer interfaces.In this paper,the construction of brain function network and the pattern recognition method based on brain function network for motor imagery EEG signals are studied.Firstly,the research background and purpose significance of motor imaging EEG signals are expounded.Then it summarizes the research status of brain function network and motion imaging EEG signal pattern recognition technology at domestic and foreign,and puts forward some problems that need to be solved.Finally,research on these issues was carried out.This article has completed some research work:(1)EEG signal preprocessing:First,the collected EEG signals are filtered to obtain the frequency band components we are interested in,and filter out high frequency noise and unwanted band components.Then do the artifact processing to remove the interference of ECG and EEG.Finally,we get pure motor imagery EEG signals.Pre-processing provides a good and effective motor-imagining EEG signal for the construction analysis of the brain function network and the pattern recognition of EEG signals.(2)Construction of brain function network:This paper introduces a multivariate Granger causality analysis method that reflects the interaction between multiple variables in a cluster,and optimizes the construction of motion imaging causal networks,and overcomes the lack of Granger causality that can only reflect the interaction between two variables.Using the same method to extract and classify causal networks based on multivariate Granger causality and based on Granger causality,It was found that there is better classification accuracy base on multivariate Granger causality than Granger.Therefore,the causal network constructed based on multivariate Granger causality can more accurately reflect the state of motor imaging neural activity.(3)Identification of multi-class motion imaging EEG signals:This paper proposes a multi-class motion imaging EEG signal recognition method based on convolutional neural network and causal network.The causal network covers multiple functional brain regions of the brain,reflecting the activity of the brain more than a single electrode or several electrodes.Using causal networks to identify multiple types of motor imagery EEG signals,first construct a causal network using multivariate Granger causality,then use causal network matrixes to train the convolutional neural network and determine the model of the convolutional neural network.Finally,the model is used to classify multiple types of motor imagery EEG signals,and better accuracy is obtained.This method provides a new idea for the recognition of multi-class motion imaging EEG signals.