A Flexible Data Set for Radar-Based Gesture Recognition

Gesture recognition is a simple and intuitive way of human-machine interaction that is increasingly used to control devices in a variety of applications. Often, the gestures are captured by a radar sensor due to its flexible operation, robustness to challenging lighting conditions, and high privacy capability. Nonetheless, radar-based data sets to train classification structures are still rare. A broad and flexible data set for radar-based gesture recognition of 7 gestures using a fast chirp frequency-modulated continuous wave (FC-FMCW) radar is presented and made publicly available. A descriptive demonstrator is built up and trained to showcase possible usage.