共空间模式结合卷积神经网络的脑电信号分类

EEG signal classification based on typical spatial pattern and convolutional neural network

为了更好解码大脑的意图和思想并作出准确识别,结合共空间模式(CSP)和卷积神经网络(CNN)算法,对脑电数据进行空间滤波处理和时空域上的特征提取。脑电信号是一种具有时空特性的信号,设计了一种CNN网络结构来进行运动想象信号的分类。为了提高分类准确率,对CSP算法中m参数进行了选择。最后,将该算法应用于公共数据集,建立分类模型并和单独使用CNN算法作对比,实验结果表明结合算法有效地提高了分类准确率,能更好地进行分类识别。

In order to better decode the brain’s intentions and thoughts and make accurate recognition,combined with the typical spatial pattern(CSP)and convolutional neural network(CNN)algorithms,the EEG data is spatially filtered and extracted in the spatio-temporal domain.EEG signal is a signal with spatio-temporal characteristics.A CNN network structure is designed to classify motor imagery signals.In order to improve the classification accuracy,the m parameter in the CSP algorithm is selected.Finally,the algorithm is applied to a public data set.A classification model is established and compared with the CNN algorithm alone.The experimental results show that the combined algorithm effectively improves classification accuracy and can better perform classification recognition.