Deep-Learning-based Human Classification on mm-Wave Radar Data

The use of radar technology in the field of perception of mobile and stationary robots has increased significantly in the last recent years. More and more sensor manufacturers and system integrators are relying on the robust properties of this method for environment sensitivity. The interpretation of the measured raw data requires sophisticated signal processing to obtain an informative and interpretable result. The signal characteristics also allow classification approaches to distinguish different objects from each other, which is escpecially interesting for the detection of humans in the working area of robots. In this paper, a selection of classification techniques using deep neural networks on radar data is presented and discussed. Finally, a method for classifying radar data from a mobile robot is proposed.