Pantomime: Mid-Air Gesture Recognition with Sparse Millimeter-Wave Radar Point Clouds
summary
A mid-range FMCW radar is configured to output sparse 3D point clouds and a neural pipeline (PointNet++ per frame + LSTM temporal fusion) classifies 21 gesture classes with real-time low compute. Reported results show 95% accuracy and 99% AUC across 21 gestures from 41 participants, with per-condition drops: ~92% for medium/fast, ~85% for slow, ~64% for through-wall; robustness is tested across five indoor environments, angles, speeds, distances up to 5 m, and the dataset and models are public.
AI-written summary · about
related talks
20mInteracting with Soli: Exploring Fine-Grained Dynamic Gesture Recognition in the Radio-Frequency ...
20mAdvancing Hand Gesture Recognition with High Resolution Electrical Impedance Tomography
19mA Rapid Prototyping Approach to Synthetic Data Generation For Improved 2D Gesture Recognition
22mStrike a Pose: Gesture Recognition in JS with Machine Learning
28mStrike a Pose - Gesture recognition in JavaScript w/ ML & Arduino
10mIMUs FTW!! Building IMU-based gesture recognition!
10mA large, crowdsourced evaluation of gesture generation systems on common data: The GENEA ...
19mBackHand: Sensing Hand Gestures via Back of the Hand