Fine Hand Gesture Recognition Using D-band FMCW Radar

This paper presents a hand gesture recognition signal processing pipeline for a D-band FMCW radar system. The setup allows to accurately classify hand gestures with fine finger movements using only a single antenna. The radar system uses a wide 45 GHz bandwidth, allowing it to detect subtle hand and finger movements. A real-time implementation is achieved by extracting a multi-feature tensor from range-Doppler maps, resulting in a compact and accurate model. To overcome overfitting, a simple and effective data augmentation technique is proposed. The system successfully recognizes six challenging hand gestures with high accuracy.