Evaluating Comfort and Performance with Composite Frequency in SSVEP-BCI

Open Access
Article
Conference Proceedings
Authors: Sodai KondoHisaya Tanaka

Abstract: Steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) offer high accuracy, fast response, and multiple input options. However, the flickering stimuli used to induce SSVEP can cause discomfort and visual fatigue. Improving user comfort without degrading performance is an important challenge in practical SSVEP-BCI applications. In this study, we developed an SSVEP-BCI using composite visual stimuli that combine high- and low-frequency flickers. Low-frequency flickers typically elicit stronger SSVEP responses, while high-frequency flickers are generally more comfortable. We hypothesized that combining both could improve comfort while maintaining BCI performance. Five stimulus conditions were tested by varying the high-to-low frequency ratio: 0%: 100%, 25%: 75%, 50%: 50%, 75%: 25%, and 100%: 0%. Each participant used the SSVEP-BCI with four inputs under all five conditions. Subjective comfort was evaluated using a 6-point scale. Results showed that BCI accuracy increased with a higher proportion of low-frequency content. The mean classification accuracies for high-frequency ratios of 100%, 75%, 50%, 25%, and 0% were 61.11±1.26%, 95.56±2.79%, 95.28±0.94%, 98.61±2.78%, and 98.61±1.39%, respectively. However, even with a higher proportion of high-frequency content, performance remained at a practically usable level. In contrast, subjective comfort scores increased with a higher proportion of high-frequency content, recorded as 5.67, 4.33, 3.22, 2.00, and 2.44, respectively. These findings indicate that composite flicker stimuli can enhance comfort while preserving SSVEP-BCI performance. Adjusting the frequency ratio allows for flexible optimization depending on the application context.

Keywords: steady-state visual evoked potential, brain-computer interface, canonical correlation analysis, composite frequency

DOI: 10.54941/ahfe1006876

Cite this paper:

Downloads
12
Visits
42
Download