Research On Task State EEG Signal Recognition Method For Spatial Cognition Assessment

The 21 st century is an era of rapid development in the field of brain science.Brain science provides preparation conditions for the development of cognitive science.The training and evaluation of spatial cognitive ability is also receiving extensive attention from scholars at home and abroad.Among them,the coupling feature extraction and classification recognition of EEG data has become an effective way of spatial cognition recognition,and has achieved certain results.However,in the field of spatial cognitive EEG data recognition,the influence of other channels on the coupling between the two channels and the importance of the spatial position between the channels are less considered in the feature extraction.At the same time,it is urgent to explore more efficient classifiers.problem.This paper explores the two aspects of coupling feature extraction and deep learning.It is proposed to improve the conditional mutual information,and then merge with the multi-spectral image transformation method.Based on this,a new convolutional neural network method is proposed.Firstly,Firstly,an EEG feature extraction method based on Multivariate Conditional Mutual Information-Multi-Spectral Image(MCMI-MSI)is proposed..For the conditional mutual information method,only the coupling strength and direction between the two channels can be calculated.In this paper,the multi-dimensional conditional mutual information(MCMI)method is proposed,and the coupling strength and direction of the two channels can be calculated under the influence of other channels.Then,the multi-spectral image transformation is performed on the coupling characteristics under the multi-band combination,which effectively compensates for the shortcomings of the existing methods.Then,a classification method of EEG signals based on multi-scale high-density convolutional neural networks is proposed.Aiming at the problem that the classic convolutional neural network can not obtain the ideal diversified features,multi-scale convolution kernel is used to extract the features of different fineness;at the same time,feature reuse is realized,the efficiency is improved,and the DenseNet method isintroduced;The gradient descent method is applied to the optimization process of multi-view convolutional neural networks.Finally,the task-state EEG signals tested before and after spatial cognitive training were used as experimental data by seven subjects recruited by the laboratory.The two new methods proposed above were compared with the existing feature extraction and classification methods.The results show that the two proposed algorithms are feasible,effective and accurate,and can be effectively applied to the recognition of space cognitive ability EEG signals.