Distribution Based Learning Network for Motor Imagery Electroencephalogram Classification

Abstract

The low Signal Noise Ratio (SNR) and nonstationarity of electroencephalogram (EEG) signals affect the classification accuracy badly in motor imagery electroencephalogram (MI-EEG) classification. In this paper, a Distribution Based Learning (DBL) framework based on deep learning is proposed to improve the accuracy. Firstly, the framework uses modified multi band Common Spatial Pattern (CSP) algorithm to pre-process the raw EEG signals. Secondly, a Distribution Based Learning Network (DBLN) is utilized to divide the dataset into two parts. After that, a two-step distribution based learning and testing strategy are conducted on the two parts separately. Experimental results on BCI Competition IV Dataset 2b indicate that accuracy of DBL is 3.84 % higher than the state-of-the-art, which proves the effectiveness of the algorithm.