Velocities in Human Hand Gestures for Radar-based Gesture Recognition Applications

Gesture recognition is a simple and intuitive way of human-machine interaction that is used more and more often to control devices in a variety of applications. Often, the gestures are captured by a radar sensor due to its robustness to challenging lighting conditions and high privacy capability. To tailor the sensor parametrization to the use in gesture recognition, the expectable velocities, amongst other parameters, have to be known. An investigation with a set of 15 defined gestures plus one individually chosen movement and 25 participants was carried out to find the main dependencies for these velocities and give an outline on how to approach radar parametrization concerning velocities for gesture recognition applications. The gesture set and setup of a certain application were identified to be the most relevant design choices, from which the expectable velocities can be determined with just a few test participants.