Human Activity Classification Using mm-Wave FMCW Radar by Improved Representation Learning

The paper proposes a novel Euclidean distance softmax layer for radar-based human activity classification. The method aims to overcome the angular dependency of classical softmax approaches. Through the freedoms thus gained, the activity classes can be distributed freely within the entire embedded feature space, due to which the dimension of the embeddings and the whole neural network size can be reduced. The performance of our novel deep learning architecture is evaluated for 60 GHz mm-wave radar sensor-based human activity classification. The results show that the proposed approach increases the robustness against random and unknown movements compared to state-of-art representation learning techniques.