A Lightweight Transformer for Pain Recognition from Brain Activity

In three linesA lightweight transformer fuses multiple fNIRS representations through unified tokenization for pain recognition from brain activity. The model preserves spatial, temporal, and time-frequency characteristics on the AI4Pain dataset, achieving competitive performance with low computational cost suitable for real-time GPU/CPU inference.

Summary generated by Claude — human-verified