Classification of Motor Imagery EEG Signals Using Deep Learning

Brain-computer interface (BCI) is a direct communication between the brain and the computer. The BCI system using an electroencephalogram (EEG), a non-invasive system, is used primarily for its high portability and practicality. This work aims to achieve the best approach to analyzing EEG signals by comparing both methods: the short-time Fourier transform (STFT) and the continuous wavelet transform (CWT). Both of these methods allow effective characterization of the signal in both time and frequency domains. Following the transformation through TFR, MI-based EEG signals are displayed as images. These images are then fed into a convolutional neural network (CNN) for feature extraction and classification. The assessment was performed using the BCI Competition 2003 dataset which includes two motor imagery classes from only one subject. Our findings demonstrated the CWT approach superior performance relative to the proposed STFT approach. In addition, the results show that the method using continuous wavelet transform with convolutional neural network can be comparable or better than the other state-of-the-art approaches.