A Hybrid Brain-Computer Interface using Extreme Learning Machines for Motor Intention Detection

This work proposes a hybrid brain-computer interface (HBCI) using Bayesian method to fuse information from electroencephalography (EEG) and electromyography signals (EMG) for decoding movement intention. For EEG signal feature extraction, Riemannian covariance matrices are computed, whereas other three features into the time and frequency domain are extracted from each EMG channel, such as autoregressive models, signal slope changes, and zero crossing. In this approach, the Bayesian fusion method is used to combine predictions from Extreme Learning Machines (ELMs) classifiers. Our proposed HBCI obtained average accuracy of 96.26%, recall of 93.97, kappa of 0.78, and false positive rate of 3.31%, outperforming solution only based on EEG or EMG, suggesting its potential for neuro-rehabilitation applications.