Cross-Subject Motor Imagery EEG Classification using Deep Learning

Brain Computer Interface (BCI) has gained a lot of attention recently due to its ability to assist people with neurological disabilities. Motor Imagery (MI) Electroencephalogram (EEG) signals are useful in BCI systems due to their high temporal resolution, which facilitates real time performance. One major limitation of BCI systems is the difficulty in obtaining generalized models that are independent of intra-subject and cross-subject variability. In this work, an approach that uses an Autoencoder for feature extraction followed by a Convolutional Neural Network (CNN) classifier is employed for improving the cross-subject generalization. The proposed method is validated using the Physionet EEG Motor Movement/Imagery Dataset, achieving an accuracy as high as 86.7% for cross-subject analysis.