"Traditional SSVEP stimuli are essentially high‑contrast brightness modulations that strongly activate the occipital visual cortex, but they are tiring," explains Professor Gao. "We asked whether we could use a more natural stimulus – rapidly changing Chinese characters – to evoke robust brain responses while being gentler on the eyes."
In the new paradigm, a sequence of Chinese characters is presented at a fixed frequency (e.g., 6 Hz), with each character appearing at the rising edge of a square wave. Unlike brightness flicker, the characters themselves carry orthographic information, engaging not only the occipital region but also the occipitotemporal cortex – areas involved in visual word processing. "This recruits a broader network of brain regions, and we hypothesize that this distributed activation may be less fatiguing," adds Professor Wang.
A frequency‑sweep experiment (3–12 Hz) revealed two distinct response patterns. At lower frequencies (3–8 Hz), the strongest EEG responses were over bilateral occipitotemporal scalp. At higher frequencies (>10 Hz), the response shifted to a more occipital‑dominant pattern. "This frequency‑dependent topography gives us flexibility in designing stimuli for different users or applications," notes Professor Gao.
Crucially, the team compared text sequence stimulation with conventional brightness flicker under varying stimulus sizes (50, 100, 150) and luminance levels (30, 127, 255). While brightness‑evoked responses were highly sensitive to both parameters – SNR variability up to 9 dB – text‑evoked responses remained stable (variability only 1.65–4.50 dB). "The text paradigm is much less affected by how large or bright the stimulus is," says lead author Xiaoyang Li. "This makes it particularly suitable for small, discreet visual interfaces."
Based on these findings, the team developed a 40‑target speller using joint frequency‑phase modulation (frequencies 3.4–11.2 Hz, phase interval 0.35π). Optimizing the decoding algorithm – task‑discriminant component analysis (TDCA) – they achieved an online average information transfer rate (ITR) of 235.12 bits/min ± 30.12 across 14 subjects, with the best user reaching 266 bits/min (100% accuracy). Notably, the stimulus duration was only 0.7 seconds, demonstrating true high‑speed operation.
"This performance is comparable to the best conventional high‑frequency SSVEP spellers, but with far less discomfort," emphasizes Professor Wang. In the offline frequency‑sweep, the TDCA algorithm achieved >90% accuracy at 0.7 s, and the best subject reached 96% accuracy at just 0.3 s (ITR 365 bits/min).
Visual comfort was assessed using two questionnaires: a custom 5‑point comfort scale and the NASA Task Load Index (NASA‑TLX). Across all measures, text sequence stimulation was rated significantly better than conventional brightness flicker. Users reported less flicker perception, higher preference, and lower physical demand. "The brightness changes during character transitions resemble natural reading, which may be why users find it less straining," explains Li.
Even across the full frequency range (1–20 Hz), text stimulation consistently outperformed brightness flicker in comfort and preference, with significant differences at most frequencies. The only exception was at very low (1 Hz) and a narrow band around 13 Hz, where text was slightly less comfortable than the optimal 8 Hz brightness flicker – but still far more comfortable than traditional high‑contrast stimuli.
The study provides neurophysiological evidence for the paradigm's advantages. Source localization and topographic analyses confirmed that text stimuli activate both occipital and occipitotemporal regions, particularly the left posterior temporal area – a key hub for visual word form processing. This ventral stream involvement may explain the robustness to size/luminance changes and the reduced fatigue, because ventral stream responses are known to be more invariant and less sensitive to low‑level physical parameters than dorsal stream or pure brightness responses.
Because text sequence stimulation works well with smaller visual areas and is less dependent on precise luminance control, it could be integrated into more discreet, wearable BCI systems – for example, using electrodes placed behind the ears or on the forehead. "This opens the door to BCIs that are not only fast but also comfortable enough for daily use," says Professor Gao.
Future work will optimize sequence design, reduce calibration time with transfer learning, and test long‑term usability in patient populations and realistic spelling scenarios. "We believe text sequence stimulation represents a new direction for high‑speed, user‑friendly visual BCIs," concludes Professor Wang.
Authors of the paper include Xiaoyang Li, Shaojie Zhang, Yonghao Song, Shangen Zhang, Xiaogang Chen, Yijun Wang, and Xiaorong Gao.
This work was supported by the National Key Research and Development Program of China under Grants 2023YFF1203702 and 2022YFC3602803; in part by the National Natural Science Foundation of China under Grants 62071447 and U2241208; in part by the Key Research and Development Program of Ningxia Hui Autonomous Region under Grant 2023BEG02063; and in part by Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (No. 2024-JKCS-26).
The paper "Text Sequence Stimulation for High-Speed and Comfortable SSVEP-BCI" was published in the journal Cyborg and Bionic Systems on Jun. 15, 2026, at DOI: 10.34133/cbsystems.0612.
