Four Novel Motion Paradigms Based on Steady-State Motion Visual Evoked Potential.

Abstract

The purpose of this paper was to study the applicability of paradigms with motion forms for use in a brain-computer interface (BCI). We examined the performances of different paradigms and evaluated the stimulus effects. We designed four novel stimulus paradigms based on basic motion modes: swing, rotation, spiral, and radial contraction-expansion. Canonical correlation analysis (CCA) was used to analyze the accuracy. Additionally, we optimized CCA template signal harmonic combinations for the different motion paradigms. The spiral motion paradigm exhibited the highest average information transfer rate (ITR) and recognition accuracy (41.24 bit/min^-1/95.33%), and the average ITRs and recognition accuracies were lowest for the rotation motion paradigm (31.89 bit/min^-1 /80.89%) and the radial contraction-expansion motion paradigm (32.62 bit/min^-1 /80.72%) because they include fewer harmonic components. Any stimulus paradigms with periodic motion can induce steady-state motion visual evoked potentials (SSMVEPs), but the SSMVEP harmonic components induced by different motion modes differed significantly. The spiral motion paradigm was more suitable for BCI applications. This study is an important extension to the existing SSMVEP-based BCI literature, and provides new insight to enable future design of the BCI paradigms.