DEEP CONVOLUTIONAL NEURAL NETWORK FOR MOTOR IMAGERY EEG SIGNALS CLASSIFICATION
Classification of electroencephalography (EEG) signals is a cornerstone in building motor-imagery based Brain-computer interface (BCI) systems. EEG signals are noisy and differ from one subject to another and even for the same subject among different trials, and this is why designing a general classification model is not an easy mission. Convolutional Neural Networks (CNN) approach is dominant in computer vision and image classification, so we followed a new trend in EEG signals classification in which these signals are transformed into images, and thus classifying such signals become an image classification problem.The motor imagery EEG signals activity is mainly in the Mu [8-13 Hz] and Beta [13-30 Hz] bands, so we band passed the EEG signals accordingly. We used the Physionet dataset for EEG motor movement/imagery tasks which consists of 109 subjects (we used 103 subjects data). The motor imagery EEG signals were transformed into 2-D images (with 2 channels, one for Mu band and the other for Beta band) using the azimuthal projection and Clough-Tocher algorithm for interpolation, and these input images are fed to a deep CNN model to classify 5 different classes (4 motor imagery tasks and one rest).Our results were promising (73.7% peak accuracy). This model is the first step for building a general purpose model which can work effectively over different datasets.