How to build a fast and accurate code-modulated brain-computer interface

Abstract Objective. In the last decade, advent of code-modulated brain-computer interfaces (BCIs) has allowed implementation systems with high information transfer rates (ITRs) and increased possible practicality such interfaces. this paper, we evaluate effect different numbers targets in stimulus display, modulation sequences generators, signal processing algorithms on accuracy ITR BCIs. Approach. We use both real simulated electroencephalographic (EEG) data, to these parameters methods. Then, compared numerous setups assess their performance identify best configurations. also evaluated dependability our evaluation approach. Main results. Our results show that Golay, almost perfect, deBruijn sequence-based visual modulations provide results, significantly outperforming commonly used m-sequences all cases. conclude artificial neural network offer pipeline for type BCI, achieving a maximum classification 94.7% EEG data while obtaining 127.2 bits min ?1 64-target system. Significance. framework demonstrated previously unattainable flexibility convenience staying reasonably realistic. Furthermore, findings suggest several new considerations which can be guide further code-based BCI development.