A Reliable Brain-Computer Interface Based on SSVEP Using Online Recursive Independent Component Analysis

Abstract

This paper presents a reliable brain-computer interface (BCI) based on a steady-state visually evoked potential (SSEVP) method using online recursive independent component analysis (ORICA) with denoising. The proposed system includes a visual stimulator, a front-end data acquisition module, an ORICA module, a power spectrum density (PSD)-based noise channel detection module, a denoising module, and an EEG reconstruction module, and a detection module using canonical correlation analysis (CCA). The system with the proposed PSD-based denoising mechanism is simulated using test patterns of 9-Hz and 10-Hz SSEVP-based EEG raw data stream with an 8-second sliding window length with a 1-second step size under the condition of 128 Hz sampling rate. The accuracy of the detection is approximately 88% and 95% hit rate for 9-Hz and 10-Hz test patterns, respectively.