Non-binary m-sequences for more comfortable brain–computer interfaces based on c-VEPs

Abstract

Background and Objectives:Code-modulated visual evoked potentials (c-VEPs) have marked a milestone in the scientific literature due to their ability to achieve reliable, high-speed brain–computer interfaces (BCIs) for communication and control. Generally, these expert systems rely on encoding each command with shifted versions of binary pseudorandom sequences, i.e., flashing black and white targets according to the shifted code. Despite the excellent results in terms of accuracy and selection time, these high-contrast stimuli cause eyestrain for some users. In this work, we propose the use of non-binary p-ary m-sequences, whose levels are encoded with different shades of gray, as a more pleasant alternative than traditional binary codes. The performance and visual fatigue of these p-ary m-sequences, as well as their ability to provide reliable c-VEP-based BCIs, are analyzed for the first time. Methods:Five different m-sequences were evaluated with 16 healthy participants, following the circular shifting paradigm: base 2 (63 bits), base 3 (80 bits), base 5 (124 bits), base 7 (48 bits), and base 11 (120 bits). Signal processing consisted of a 3-filter bank (1–60 Hz, 12–60 Hz and 30–60 Hz), followed by a canonical correlation analysis. Raster latency correction and artifact rejection approaches were also applied to compute command templates. For each m-sequence, users performed a 30-trial calibration stage, followed by an online spelling task of 32 trials. In addition, qualitative measures regarding visual fatigue and satisfaction were collected. Results:Users were able to achieve an average accuracy of over 98% for all p-ary m-sequences. The differences between m-sequences were not significant in terms of accuracy, but they were in terms of visual fatigue. The higher the base, the less eyestrain perceived by users for both presentation rates of 60 Hz and 120 Hz. All p-ary m-sequences were also significantly less annoying when displayed at 120 Hz compared to 60 Hz. Conclusion:Results suggest that all p-ary m-sequences are suitable for achieving high speed and high accuracy in c-VEP-based BCIs, reducing the visual fatigue as the base increases, without degrading system performance. It is concluded that the use of high presentation rates and non-binary m-sequences is a promising alternative to provide user-friendly c-VEP-based BCIs.