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

Abstract

Objective. Learning the structures and unknown correlations of a motor imagery electroencephalogram (MI-EEG) signal is important for its classification. It is also a major challenge to obtain good classification accuracy from the increased number of classes and increased variability from different people. In this study, a four-class MI task is investigated. Approach. An end-to-end novel hybrid deep learning scheme is developed to decode the MI task from EEG data. The proposed algorithm consists of two parts: a. A one-versus-rest filter bank common spatial pattern is adopted to preprocess and pre-extract the features of the four-class MI signal. b. A hybrid deep network based on the convolutional neural network and long-term short-term memory network is proposed to extract and learn the spatial and temporal features of the MI signal simultaneously. Main results. The main contribution of this paper is to propose a hybrid deep network framework to improve the classification accuracy of the 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 subjects to form one model. Significance. The classification performance obtained by the proposed algorithm on brain–computer interface (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 experimental results illustrate that the shared neural network has satisfactory accuracy. Thus, the proposed algorithm could be of great interest for real-life BCIs.

Single-Trial EEG Classification with EEGNet and Neural Structured Learning for Improving BCI Performance

“…Though it is to be noted that in the original paper the EEG had undergone feature engineering through the use of common spatial patterns while in our case the EEG itself was used as a feature. With regards to the multi-class performance, current state-of-the-art has higher reported accuracy of 83% by Zhang et al [42] and 75.77% by Amin et al [43] as compared to our best accuracy of 70.68% on the same dataset. This being the first use of neural structured learning for EEG classification it is expected that with further parameter tuning and better implementation of structured learning that complements well with the base model, the model accuracy can be further be improved.…”

A Parallel Multiscale Filter Bank Convolutional Neural Networks for Motor Imagery EEG Classification

“…Both of them yielded higher accuracies compared with FBCSP. Hauke et al used a simplified CNN model to validate that a DL model was effective in transfer learning tasks for recordings from 109 subjects (Goldberger et al, 2000) without any preprocessing (Dose et al, 2018).…”

An Approach of One-vs-Rest Filter Bank Common Spatial Pattern and Spiking Neural Networks for Multiple Motor Imagery Decoding

“…Compared with other methods, this method obtained the best recognition accuracy [20]. Zhang et al used OVR-FBCSP combined with CNN+LSTM method for MI recognition [53]. Islam et al used TSM to find the precise frequency band associated with the MI mission for MI recognition [41].…”