Pascal GerhardsInfineon Technologies Dresden GmbH & Co. KG
KI-ASIC_Radar-based gesture recognition with signal processing in spiking neural networks
Spiking neural networks (SNN) are third generation neural networks with large potential in spatio-temporal tasks regarding latency and energy consumption. In this work, we train a SNN for radar-based gesture recognition as input to a game. We show how to simulate and directly train the SNN in TensorFlow with surrogate gradients on range-doppler- and range-angle-maps. We also show that the performance of SNN is on par with classical deep neural networks (DNN).