Comparative study of methods for the frequency recognition of Steady state visual evoked potentials on experienced and naive users on Brain-Computer Interface

Estudio comparativo de métodos para el reconocimiento frecuencial de potenciales evocados visuales en estado estacionario en usuarios con y sin experiencia en Interfaces Cerebro-Computador

One of the advantages reported in the literature about of Steady State Visual Evoked Potentials (SSVEP) over other paradigms for use in a Brain-Computer Interface (BCI) is that it is possible to obtain an adequate classification of commands in users without BCI experience (naives). The present article focuses on quantifying the variation in the performance of a BCI-SSVEP for users with experience (group 1) and without experience (group 2). Two of the state-of-the-art methods for SSVEP recognition were used: Canonical Correlation Analysis (CCA) and Canonical Correlation Analysis using Filter Bank (FBCCA). For this study, a database referenced in other studies (benchmark) is used, it is composed of 40 different visual stimuli, the analysis was performed with 16 subjects (8 each group). Classification Percentage and ITR were used as rating metrics to evaluate the performance of methods, were evaluated in 4-time windows: 0.5 s, 1 s, 1.5 s and 2 s. In addition, a statistical significance analysis was performed. As result, it was obtained that the best SSVEP recognition method correspond to the FBCCA with average percentage of classification of 93.31% for group 1 and 89.22% for group 2, and ITR of 169.85 bits/minute for group 1 and 142.76 bits/minute for group 2, in this last case evaluated in a time window of 1.5 s. Finally, the statistical analysis with a significance of 5%, allows to conclude that the experience in BCI has a low influence on the performance of a BCI-SSVEP system.