Gesture Recognition

Gesture recognition studies how to detect and interpret movements, from hand signs and subtle social cues to commands for robots, XR systems, and assistive devices. Common approaches use CNNs, temporal models such as transformers and recurrent networks, pose or skeleton graphs, and fusion of video with inertial, EMG, radar, or Wi-Fi signals; benchmarks include Jester, NVGesture, and iMiGUE. Recent work also explores self-supervised learning, synthetic data, and compact models for real-time operation on phones, wearables, and robots. A central challenge is that strong results on familiar users or controlled data often do not transfer to new people and settings, while continuous, safety-critical use also demands low latency and reliable handling of uncertainty and false triggers.