A simplified hybrid EEG-fNIRS Brain-Computer Interface for motor task classification

Hybrid Brain-Computer Interfaces (BCI) has shown great promise for neuro-prosthetics and assistive devices in the field of rehabilitation. However, the complexity involved with the system design and time cost for classification of motor tasks is a core problem when we step into clinical applications. To help address this problem, simultaneous measurements of Electroencephalography (EEG) and functional Near-Infrared Spectroscopy (fNIRS) signals were carried out on three healthy volunteers under a left-right hand grasping paradigm. In this work, we applied a wavelet-based method to extract the wavelet approximation coefficients from EEG signals and we employed the slope information to discriminate the concentration change of Oxygenated hemoglobin (HbO) during left-right hand grasping tasks. To maximize the valuable information implied in two modalities, we proposed an approach based on principle component analysis (PCA) to integrate the features of fNIRS and EEG signals. Two classifiers, including support vector machine (SVM) and linear discriminant analysis (LDA) were applied to identify and estimate the control signals associated with left-right hand grasping tasks. The present experimental result demonstrates that the complement of EEG and fNIRS can significantly improve the classification accuracy with 3~9% on average. The reduction of dimensionality by PCA could achieve a reduction of time complexity and computational complexity with little loss in accuracy.