A novel hybrid deep learning scheme for four-class motor imagery classification.

Researchers

Journal

Modalities

Models

Abstract

Learning the structures and unknown correlations of motor imagery (MI) EEG signal is important for the classification. It is also a major challenge to obtain good classification accuracy from increased number of classes and increased variability from different people. In this study, four-class MI task is investigated. An end-to-end novel hybrid deep learning scheme is developed to decode MI task from EEG data. The proposed algorithm consists of two parts: \emph{a}. One-versus-rest filter bank common spatial pattern (OVR-FBCSP) is adopted to preprocess and pre-extract the features of four-class MI signal. \emph{b}. A hybrid deep network based on CNN and LSTM network is proposed to extract and learn the spatial and temporal features of MI signal simultaneously. The main contribution of this paper is to propose a hybrid deep network framework to improve classification accuracy of four-class MI-EEG signal. The hybrid deep network is a subject-independent shared neural network which means it can be trained by using the training data from all the subjects to form one model. The classification performance obtained by the proposed algorithm on BCI competition IV dataset 2a in terms of accuracy is 83% and Cohen’s kappa value is 0.80. Finally, the shared hybrid deep network is evaluated by every subject respectively, and the experiment results illustrate that the shared neural network get a satisfactory accuracy. Thus, the proposed algorithm could be of great interest for real-life brain-computer interfaces (BCIs).
© 2019 IOP Publishing Ltd.