A Novel Hybrid BCI System Combining Single-Channel SSVEP and PLR to Improve Classification Accuracy and ITR

Abstract Steady-state visual evoked potential (SSVEP)-based brain–computer interfaces (BCIs) have been widely studied because they provide high classification accuracy and information transfer rate (ITR) without requiring user training. To further enhance BCI performance, this study proposes a novel hybrid BCI that integrates single-channel SSVEP with the pupillary light reflex (PLR). Twelve healthy subjects participated in experiments involving three paradigms: SSVEP, PLR, and hybrid. Each subject completed a total of 90 trials (three blocks) for each of the four classes in every paradigm. The experimental protocols were identical across paradigms, except for differences in visual stimuli. The hybrid paradigm achieved an accuracy (ITR) of 95.70% (25.54 bpm) for four-class classification using a data length of 4 s across all subjects, representing increases of 8.54% (6.34 bpm) and 12.92% (8.91 bpm) compared with the SSVEP and PLR paradigms, respectively. In the poor performer group (SSVEP accuracy below 90%), the hybrid paradigm achieved an average accuracy (ITR) of 93.06% (23.03 bpm), showing an improvement of 11.67% (7.54 bpm) compared with the SSVEP paradigm (81.39% and 15.49 bpm). As the number of EEG channels increased from one to ten electrodes, the classification accuracy of the SSVEP paradigm improved from 87.15% to 91.28% (an increase of 4.13%). In contrast, the hybrid paradigm exhibited minimal change, with only a 0.55% increase (from 95.70% to 96.25%). These results suggest that the proposed hybrid BCI can serve as an effective approach to enhance overall BCI performance, which may contribute to the practical and widespread adoption of BCIs by improving usability and user experience.