Employing Deep Learning and Discrete Wavelet Transform Approach to Classify Motor Imagery Based Brain Computer Interface System

In the last few years, Brain Computer Interfaces (BCI) attempted the attention of many researchers. In Motor Imagery (MI)-BCI, central nervous system directly connected to a computer or an automation system. Characteristics of the electromyographic (EEG) signals are utilized in MI-BCI systems. Various techniques have been proposed to extract EEG signal characteristics during recent years. The main objective of this research is to employ an efficient deep learning approach to extract the features of EEG signals using composition of convolutional Neural Network and discrete wavelet transform utilized in the BCI system. The deep learning approach presented in this study has rarely been explored to employ for EEG features extraction. The simulation study indicates that the presented method carry out remarkable accuracy and high performance compared with conventional approaches such as support vector machine and artificial Neural Network methods and give a powerful indicative decision making aid to assist physicians in the treatment of the right and left-hand features for real time motor imagery classification system. Furthermore, the most advantages of employing the proposed method are to eliminate the feature selection level and reducing the processing cost significantly.