Recognition of single upper limb motor imagery tasks from EEG using multi-branch fusion convolutional neural network

• DOI: 10.3389/fnins.2023.1129049
• Publication Date: 2023-02-22T09:33:10Z
• Authors: Rui Zhang; Yadi Chen; Zongxin Xu; Lipeng Zhang; Yuxia Hu; Mingming Chen
• Keywords: Motor Imagery; Neurorehabilitation; Binary classification

Motor imagery-based brain-computer interfaces (MI-BCI) have important application values in the field of neurorehabilitation and robot control. At present, MI-BCI mostly use bilateral upper limb motor tasks, but there are relatively few studies on single MI tasks. In this work, we conducted recognition imagery EEG signals right proposed a multi-branch fusion convolutional neural network (MF-CNN) for learning features raw as well two-dimensional time-frequency maps at same time. The dataset used study contained three types tasks: extending arm, rotating wrist, grasping object, 25 subjects were included. binary classification experiment between object arm-extending MF-CNN achieved an average accuracy 78.52% kappa value 0.57. When all tasks classification, 57.06% 0.36, respectively. comparison results showed that performance is higher than CNN branch algorithms both binary-class three-class classification. conclusion, makes full time-domain frequency-domain EEG, can improve decoding it contributes to function rehabilitation training after stroke.