EEG Motor Imagery Classification Based on Multi-spatial Convolutional Neural Network

EEG motor imagery classification has important implications for the development of brain-computer interfaces. Unfortunately, how to accurately and comprehensively utilize the feature information contained in EEG motor imagery signals to further improve the classification performance is still a challenge. To solve this problem, this paper proposes an EEG motion imagery classification model based on multiple spatial convolution kernels. The model consists of spatial convolution and temporal convolution to simultaneously extract the feature expressions of EEG signals in different spaces. The experimental results show that the algorithm proposed in this paper achieves better classification accuracy than most existing algorithms in multiple data sets, which reflects the superiority of the algorithm. The work in this paper will advance the field of EEG motor imagery.