BCI Speller on Smartphone Device with Diminutive-Sized Visual Stimuli

Abstract

In real-world BCI applications, small-sized and low-impact stimuli are more appropriate for smart devices. However, diminishing the stimuli intensity leads to a reduction of P300 amplitude, causing lower system performance. The purpose of this study is to propose a state-of-the-art BCI speller where diminutive (less than 1 mm) visual stimuli were implemented in a smartphone interface. To boost the task-relevant brain components, participants performed a certain mental task according to the given cue signs. Additionally, we applied a data-driven optimization approach to represent the user-specific spatial-temporal features. The results showed 96.8% of spelling accuracy with a maximum ITR of 31.6 [bits/min], which is comparable or even superior to conventional speller systems. Our study demonstrated the feasibility to create more reliable and practical BCI spelling systems in the future.