Binary Convolutional Neural Network for Brain Computer Interfaces

Abstract

Recent years, there is increasing interest in using deep neural network for EEG analysis. In this paper, we have innovatively applied the data augmentation commonly used in CNN image classification to EEG signals classification, which can improve classification performance: motor imagery (MI) tasks (left hand and right hand) accuracy rate can reach 93.1%. Then, we innovatively use BCNN (Binary Convolutional Neural Network) to analyze the MI tasks (left hand and right hand movement). BCNN can reduce power consumption and reduce the time complexity by 60% in dedicated hardware. BCNN is also easy to map to general hardware. In addition, we use BCNN as a classification module and use Emotiv EPOC as collection device to build BCI platform.