Mutual Information-Based Electrode Selection Extended With Prior Knowledge For Use in Brain-Computer Interfacing

Brain-computer interfacing (BCI) benefits from a multi-electrode approach, certainly when relying on non-invasive recording techniques such as EEG. However, as the set of electrodes that maximize BCI performance is paradigm-and subject dependent, selecting the optimal electrode set from the vast number of possible electrode subsets is challenging. In this work, an existing electrode selection algorithm based Mutual Information (MI) is extended with prior knowledge of the paradigm under investigation, and applied towards a P300-based BCI. The results show that the proposed algorithm is at par with the original MI electrode selection and outperforms a traditional greedy forward electrode selection approach, but requires significantly less time to converge.