A Novel Hybrid Brain–Computer Interface for EEG and EMG Based on an Ensemble Learning Approach for Motor Imagery Classification Tasks

The acquisition, analysis, and measurement of biological signals are fundamental methods for exploring the current state and predicting the future state of organisms. Among them, the processing of human electroencephalogram (EEG) signals in motor imagery (MI) tasks is a current frontier research direction in the field of brain-computer interfaces (BCIs). However, EEG-based BCIs are often susceptible to noise interference, limiting their applicability. To address these challenges, our goal is to develop a novel and efficient hybrid BCI model that integrates EEG and electromyogram (EMG) signals, with the aim of improving both the accuracy and the range of MI, as well as constructing a corresponding dataset. The method involves the use of the bilinear submanifold learning (BSML) algorithm and the minimum distance to the bilinear submanifold mean (MDSM) algorithm for EEG signals, sliding window, L1 regularization, and the support vector machine (SVM) algorithm for EMG signals, and transforming the classification problem into a regression problem and constructed ensemble learning model, multiobjective BSML-MDSM FE-SVM (MOBF). Furthermore, extensive experiments on both homemade and publicly available datasets, leave-one-out cross-validation (LOOCV) were applied to assess the model’s performance. The results demonstrate that the proposed MOBF model outperforms the state-of-the-art (SOTA) methods, achieving improvements in MI accuracy of 2.4% and 1.1% on the homemade and public datasets, respectively. Notably, the model also achieved the highest accuracy across different EEG frequency bands and for muscle fatigue detection.