Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module Chmurski, Mateusz Mauro, Gianfranco Santra, Avik Zubert, Mariusz Dagasan, Gökberk Edge computing Edge TPU Optimization Quantization FMCW Radar Deep learning Neural networks The increasing integration of technology in our daily lives demands the development of more convenient human–computer interaction (HCI) methods. Most of the current hand-based HCI strategies exhibit various limitations, e.g., sensibility to variable lighting conditions and limitations on the operating environment. Further, the deployment of such systems is often not performed in resource-constrained contexts. Inspired by the MobileNetV1 deep learning network, this paper presents a novel hand gesture recognition system based on frequency-modulated continuous wave (FMCW) radar, exhibiting a higher recognition accuracy in comparison to the state-of-the-art systems. First of all, the paper introduces a method to simplify radar preprocessing while preserving the main information of the performed gestures. Then, a deep neural classifier with the novel Depthwise Expansion Module based on the depthwise separable convolutions is presented. The introduced classifier is optimized and deployed on the Coral Edge TPU board. The system defines and adopts eight different hand gestures performed by five users, offering a classification accuracy of 98.13% while operating in a low-power and resource-constrained environment. 2021-11-09T11:24:59Z 2021-11-09T11:24:59Z 2021 journal article Chmurski, M.; Mauro, G.; Santra, A.; Zubert, M.; Dagasan, G. Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module. Sensors 2021, 21, 7298. https:// doi.org/10.3390/s21217298 http://hdl.handle.net/10481/71388 10.3390/s21217298 eng http://creativecommons.org/licenses/by/3.0/es/ open access Atribución 3.0 España MDPI