A User-Friendly Ssvep-Based Brain-Computer Interface Using A Time-Domain Classifier

An Luo,Thomas J. Sullivan
DOI: https://doi.org/10.1088/1741-2560/7/2/026010
IF: 5.043
2010-01-01
Journal of Neural Engineering
Abstract:We introduce a user-friendly steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) system. Single-channel EEG is recorded using a low-noise dry electrode. Compared to traditional gel-based multi-sensor EEG systems, a dry sensor proves to be more convenient, comfortable and cost effective. A hardware system was built that displays four LED light panels flashing at different frequencies and synchronizes with EEG acquisition. The visual stimuli have been carefully designed such that potential risk to photosensitive people is minimized. We describe a novel stimulus-locked inter-trace correlation (SLIC) method for SSVEP classification using EEG time-locked to stimulus onsets. We studied how the performance of the algorithm is affected by different selection of parameters. Using the SLIC method, the average light detection rate is 75.8% with very low error rates (an 8.4% false positive rate and a 1.3% misclassification rate). Compared to a traditional frequency-domain-based method, the SLIC method is more robust (resulting in less annoyance to the users) and is also suitable for irregular stimulus patterns.
What problem does this paper attempt to address?