A two-stage transformer based network for motor imagery classification

• Published in: Medical engineering & physics Vol. 128; p. 104154
• Main Authors: Chaudhary, Priyanshu, Dhankhar, Nischay, Singhal, Amit, Rana, K.P.S.
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 01.06.2024
• Subjects: Augmentation; Brain-Computer Interfaces; Classification; Deep Learning; Electroencephalography; Humans; Imagination - physiology; Motor Activity - physiology; Movement; Neural Networks, Computer; Signal Processing, Computer-Assisted; TabNet; Temporal convolution network (TCN); Two-stage architecture; Augmentation; Two-stage architecture; TabNet; Temporal convolution network (TCN); Classification
• ISSN: 1350-4533; 1873-4030; 1873-4030
• DOI: 10.1016/j.medengphy.2024.104154

•A two-stage architecture is developed for motor imagery classification.•Fusion of handcrafted features and deep learning based embeddings is employed.•Proposed method uses multi-head attention, separable temporal convolutions, TabNet.•For data augmentation, a novel channel cluster swapping is implemented.•Enhanced accuracy by approximately 3 % on the BCI datasets was obtained. Brain-computer interfaces (BCIs) are used to understand brain functioning and develop therapies for neurological and neurodegenerative disorders. Therefore, BCIs are crucial in rehabilitating motor dysfunction and advancing motor imagery applications. For motor imagery, electroencephalogram (EEG) signals are used to classify the subject's intention of moving a body part without actually moving it. This paper presents a two-stage transformer-based architecture that employs handcrafted features and deep learning techniques to enhance the classification performance on benchmarked EEG signals. Stage-1 is built on parallel convolution based EEGNet, multi-head attention, and separable temporal convolution networks for spatiotemporal feature extraction. Further, for enhanced classification, in stage-2, additional features and embeddings extracted from stage-1 are used to train TabNet. In addition, a novel channel cluster swapping data augmentation technique is also developed to handle the issue of limited samples for training deep learning architectures. The developed two-stage architecture offered an average classification accuracy of 88.5 % and 88.3 % on the BCI Competition IV-2a and IV-2b datasets, respectively, which is approximately 3.0 % superior over similar recent reported works.