Comparative study of SSVEP characteristics in mixed versus virtual reality across varying depths

TL;DR

This study compares how mixed and virtual reality affect brain signals used in brain-computer interfaces, finding that mixed reality performs better for visual comfort and signal stability.

Contribution

The study experimentally validates optimization strategies for SSVEP-based BCIs in neurorehabilitation by comparing MR and VR environments.

Findings

Mixed reality (MR) environments showed higher signal-to-noise ratio (SNR) than virtual reality (VR), especially at 1.8 m depth.

FBCCA achieved the highest SSVEP classification accuracy at 0.4 m depth in MR environments.

11.25 Hz flicker frequency elicited the highest SSVEP amplitude and SNR, making it optimal for BCI applications.

Abstract

Steady-state visually evoked potentials (SSVEP), owing to their high signal-to-noise ratio and low training cost, are widely regarded as an effective approach for constructing visually driven brain-computer interfaces (BCI), particularly in neurorehabilitation applications. However, the accommodation-vergence conflict (VAC) commonly present in mixed reality (MR) and virtual reality (VR) head-mounted displays may attenuate neural responses in the visual cortex, thereby compromising the long-term usability of such systems. This study aims to systematically evaluate the effects of MR and VR environments under different virtual depth conditions on SSVEP signal quality, classification performance, and visual comfort, providing parameter guidelines for the design of immersive visual BCIs in rehabilitation contexts. Green flickering stimuli at 7.5, 11.25, and 18 Hz were presented at three virtual depths of 0.4, 1.0, and 1.8 m. Feature extraction and classification were performed using canonical correlation analysis (CCA), Filter-Bank Canonical Correlation Analysis (FBCCA), and task-related component analysis (TRCA).The results showed a negative correlation between stimulus distance and SSVEP classification accuracy, with FBCCA achieving the highest accuracy at the 0.4 m depth (71.8% ± 33.8%). Overall, the signal-to-noise ratio (SNR) in the MR environment was higher than that in the VR environment, with the most pronounced difference observed under the 1.8 m condition, suggesting that MR is more effective in alleviating VAC and maintaining stable visual cortical responses. Among the three stimulation frequencies, 11.25 Hz elicited the highest SSVEP amplitude and SNR, indicating it as the optimal frequency band. Subjective visual fatigue assessments revealed higher scores for VR in terms of diplopia and fixation difficulty, with trends consistent with the observed SNR reduction. This study elucidates the interactive modulation effects of virtual depth, display modality, and flicker frequency on SSVEP, and demonstrates that MR outperforms VR in terms of signal stability, visual comfort, and potential rehabilitation usability. The derived parameters provide experimentally validated optimization strategies for stimulus depth and frequency in vision-based attention training, spatial orientation training, upper-limb interactive tasks, and immersive feedback systems in neurorehabilitation, thereby contributing to improved long-term adherence and clinical translational value of future rehabilitation BCI.

Figures