A SSVEP-Based BrainCComputer Interface With Low-Pixel Density of Stimuli
Jiayuan Meng,Hui Liu,Qiaoyi Wu,Hongzhan Zhou,Wenqiang Shi,Lin Meng,Minpeng Xu,Dong Ming
DOI: https://doi.org/10.1109/TNSRE.2023.3328917
IF: 4.9
2023-01-01
IEEE Transactions on Neural Systems and Rehabilitation Engineering
Abstract:The brain-computer interface (BCI) based on the steady-state visual evoked potential (SSVEP) has drawn widespread attention due to its high communication speed and low individual variability. However, there is still a need to enhance the comfort of SSVEP-BCI, especially considering the assurance of its effectiveness. This study aims to achieve a perfect balance between comfort and effectiveness by reducing the pixel density of SSVEP stimuli. Three experiments were conducted to determine the most suitable presentation form (flickering square vs. flickering checkerboard), pixel distribution pattern (random vs. uniform), and pixel density value (100%, 90%, 80%, 70%, 60%, 40%, 20%). Subjects' electroencephalogram (EEG) and fatigue scores were recorded, while comfort and effectiveness were measured by fatigue score and classification accuracy, respectively. The results showed that the flickering square with random pixel distribution achieved a lower fatigue score and higher accuracy. EEG responses induced by stimuli with a square-random presentation mode were then compared across various pixel densities. In both offline and online tests, the fatigue score decreased as the pixel density decreased. Strikingly, when the pixel density was above 60%, the accuracies of low-pixel-density SSVEP were all satisfactory (>90%) and showed no significant difference with that of the conventional 100%-pixel density. These results support the feasibility of using 60%-pixel density with a square-random presentation mode to improve the comfort of SSVEP-BCI, thereby promoting its practical applications in communication and control.