Transfer Learning based Motor Imagery Classification using Convolutional Neural Networks

Nowadays, classification of signals is considered as the crucial role of motor imagery brain computer interface. Moreover, deep learning approaches show acceptable performance in image recognition applications as well as speech recognition. However, practicality of the aforementioned technique is not generally deployed on motor imagery tasks. Hence, the goal of this paper is to apply convolutional neural networks to classify the motor imagery EEG signals. In addition, data augmentation along with excusive transfer learning strategy are used to overcome the problem of few trials in motor imagery tasks. On the other hand, analytical regression assessments are also applied to the raw data for mitigating the stress of EOG on EEG. Consequently, the simulation results clearly convey the contribution of the proposed algorithm via testing on BCI competition IV dataset 2b. Applying EOG artifact removal and data augmentation methods resulted in 0.07 improvement in kappa coefficient. Furthermore, using our proposed transfer learning method led to 0.06 improvement in terms of kappa coefficient.