A Wearable Multimodal Surface Electromyography Dataset for Gesture-Free and Static Gesture Recognition

Surface electromyography (sEMG) has been widely adopted for hand gesture recognition; however, existing datasets primarily focus on visible hand movements and often acquire single-modality signals, limiting model robustness in real-world scenarios. To address this gap, we present a wearable multimodal sEMG dataset covering two paradigms: static gestures (14 classes) and "gesture-free" movements (12 classes of fine-grained isometric force patterns without visible joint motion). Data were collected from 30 healthy participants over two separate days (1–3 days apart). All signals were synchronously acquired via a custom wristband and sensor-integrated glove, comprising 8-channel wrist sEMG,6-axis on the glove), 5-channel finger flexion, and 5-channel fingertip pressure, all sampled at 500 Hz. The dataset supports single-day, cross-day, and cross-subject evaluation protocols. Baseline results demonstrate that multimodal fusion substantially improves recognition accuracy. This dataset provides a new benchmark for research on robust myoelectric decoding, multimodal fusion, and wearable low-interaction human–computer interfaces.