A Hybrid BCI Using Singular Value Decomposition Values of the Fast Walsh-Hadamard Transform Coefficients

One of the main goals of a brain computer interface (BCI) is to enable a communication channel between the brain and electronic devices by converting neural activity into control commands either for devices or applications. Because of the excellent temporal resolution, low set-up cost, and noninvasive nature, BCI systems generally use electroencephalography (EEG) for an input signal. However, EEG suffers from poor spatial resolution, and it is contaminated by various external and internal artifacts, such as environmental magnetic noises and body movements. These limitations directly affect the performance of the EEG-based BCI system, and it might not work at the desired level. On the other hand, near-infrared spectroscopy (NIRS) has an advantage of relative robustness against body movements and electrical artifacts. Additionally, it is also a promising neural signal recording method which provides good spatial resolution. In this study, we particularly focused on compensating the limitations of EEG-based BCI system by adding simultaneous NIRS modality features. In order to show the effectiveness of our method, we used an open-access data set, which was recorded from 29 subjects with simultaneous EEG-NIRS system during the imagination of opening and closing either a left- or right-hand. The features were extracted by calculating the singular value decomposition values of the Fast Walsh-Hadamard transform coefficients. Afterward, the k-nearest neighbor algorithm was performed to classify the features. The performance of the proposed method was evaluated in terms of classification accuracy and kappa value metrics. The achieved results showed that combining a hybrid BCI system with EEG-NIRS modalities can enhance the performance of a BCI by 6.75% compared to the single-modality solution of EEG.