Intention Recognition Based On Eeg Signal Modeling And Spatial Filtering

In recent years,the modeling and analysis of multi-channel Electroencephalogram(EEG)has become one of the hotspots in the field of neurodynamics.However,one of the key factors affecting the development of Brain-Computer Interface(BCI)is how to obtain a multi-neuron mass model with good frequency characteristics and how to use the neuron mass model to analyze EEG signals.In addition,the non-invasive motor imagery Electroencephalography(EEG)acquisition and processing technology is widely used in clinical rehabilitation training and external equipment control.Feature extraction and pattern recognition have always been one of the key topics in EEG analysis.Therefore,the main contents of the paper are as follows:(1)Optimized multi-dynamic and multi-channel coupled Neuron Mass Model(ONMM).Based on the single-channel EEG neuron mass model,the dynamic model of neuron mass was constructed.In this paper,the factors influencing the synthesis of multi-channel EEG signals are discussed.This model not only achieves regional coupling on the channel,but also combines the interaction between cortical and thalamic neuronal groups.The results show that the ONMM model can synthesize detailed simulated EEG signals.Multi-dynamic and multi-channel EEG synthesis modeling and its influencing factors analysis is a challenging task.The optimized signal modeling scheme is helpful for the prevention,diagnosis and treatment of neurological diseases,and provides the technical basis for cognitive neuroscience research.(2)An Optimal Discriminant Hyperplane-Common Spatial Subspace Decomposition(ODH-CSSD)method for EEG feature extraction based on spatial domain filtering is proposed.The multi-dimensional EEG features were extracted from the original EEG signals using the common space subspace decomposition algorithm,and the optimal feature criterion was established to find the multi-dimensional optimal projection space.A classical method of data dimension optimization is to use the eigenvector of the lumped covariance matrix corresponding to the maximum eigenvalue.Then,the cost function is defined as the extreme value of the criterion,and n orthogonal discriminant vectors corresponding to n extreme values of the criterion are solved to construct the n-dimensional optimal feature space.Finally,multi-dimensional EEG features were projected into the n-dimensional optimal projection space,and the optimal n-dimensional EEG features were extracted.(3)Discriminant Rectangle Mixed Model(DRMM),an interpretable model based on clustering.This model can not only solve the problem that the general clustering algorithm can not provide interpretation,but also effectively overcome the time consumed by the training and testing of the model.In this paper,the multi-dimensional EEG feature matrix is optimized by(2)to improve the divisibility of the optimal EEG intention feature,and the final rectangular decision distribution rules can be obtained through the DRMM model to realize the interpretation of the clustering results.Under the condition of identifying the two-dimensional optimal features,the clustering accuracy of DRMM reaches 0.942,which is obviously better than the K-mean clustering accuracy of0.852 and the fuzzy C-mean clustering accuracy of 0.855.In conclusion,through modeling and analysis of multi-channel EEG signals,this paper studies how information is transmitted,integrated,distributed and received in multiple regions of the brain,and further explores the deep correlation of brain community structure,which has a certain significance for human understanding of brain mechanism.In the application of BCI based on motor imagery EEG,this paper proposes a method of EEG feature extraction based on spatial domain filtering to enhance EEG signal features.Based on the interpretable cluster model,the single experimental motor imagery EEG decoding was implemented for the real EEG data and the generated EEG data respectively.