A multimodal EEG-fNIRS dataset was collected from 31 participants watching three instructional videos, yielding over 4.5 million data points with expert-verified attention labels¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Simultaneous EEG and fNIRS signals were collected from 31 healthy participants during videos on convolutional neural networks, matrix factorization, and neurons and synapses. → Shixian Zhou 2026 q2 i? The dataset contains over 4.5 million data points, with experiments lasting 50.2±7.6 minutes, and labels were verified by two domain experts. → Shixian Zhou 2026 (2)

Evidence¶

Shixian Zhou 2026¶

Shixian Zhou, Minghuan Shen, Xiaolin Wen, Zijun Qiu, Yongliang Jiang, Xiangyang Wu, Fei Wu, Yong Wang, and Zhiguang Zhou. (2026). SAVVY: Student Attention Visualization for Video-based Learning Analysis. arXiv. https://arxiv.org/abs/2607.29413 doi:10.48550/arxiv.2607.29413

q2 · i? · design · r2

Brain data acquisition described in the Brain Dataset subsection. Participants watched videos on three topics; the authors note the sample of 31 "exceeds the sample sizes typically used in BCI experiments" and that the experiment was IRB-approved.

"We collected a multimodal brain dataset with simultaneous EEG and fNIRS signals from 31 healthy participants (16 male, 15 female; mean age 23.6± 1.2 years) during instructional videos on convolutional neural networks, matrix factorization, and neurons and synapses."

Shixian Zhou 2026 (2)¶

Shixian Zhou, Minghuan Shen, Xiaolin Wen, Zijun Qiu, Yongliang Jiang, Xiangyang Wu, Fei Wu, Yong Wang, and Zhiguang Zhou. (2026). SAVVY: Student Attention Visualization for Video-based Learning Analysis. arXiv. https://arxiv.org/abs/2607.29413

q2 · i? · design · r2

Dataset scale reported in the Brain Dataset subsection. Two domain experts (E3-E4) independently judged initial attention-level labels using quiz accuracy, EEG spectrograms, and fNIRS activation, with ambiguous segments discarded.

"The dataset contains over 4.5 million data points, with each experiment lasting 50.2± 7.6 minutes."

Discussion¶

Related Claims¶

• A DDQN-based attention quantification model trained on multimodal EEG-fNIRS features produces predicted attention levels at one-second temporal resolution — related
• Two case studies and expert interviews with six experts demonstrated SAVVY's effectiveness and usability for identifying attention variations and supporting video optimization — related
• The end-to-end attention-estimation pipeline meets the one-second prediction stride on an NVIDIA A100 GPU, with warm latency of 50.65 ms per window and 1.02 ms for the temporal model — related