DVSGesture Recognition with Neuromorphic Observation Space Reduction Techniques

Charles P. Rizzo, Luke McCombs, Braxton Haynie, Catherine Schuman, James Plank

Abstract: Event-based cameras and classification datasets pair nicely with neuromorphic computing. Furthermore, it is attractive from a SWaP perspective to have a fully neuromorphic pipeline from event-based camera output to classification instead of having to preprocess the camera data prior to classification. In this work, we examine how two neuromorphic observation space reduction techniques impact classification performance on the DVSGesture dataset. The two techniques can be implemented as spiking neural networks so that no preprocessing of the camera data is required, and instead, only the routing of the events to the proper input neurons is necessary.