2D-Motion Detection using SNNs with Graphene-Insulator-Graphene Memristive Synapses

The event-driven nature of spiking neural networks (SNNs) makes them biologically plausible and more energy-efficient than artificial neural networks. In this work, we demonstrate SNN-based motion detection of an object in a two-dimensional visual field. The network architecture presented here is biologically plausible and uses CMOS-based analog leaky integrate-and-fire neurons and ultra-low power RRAM synapses. We have investigated the role of visual field optimization on network performance in terms of accuracy and power cost. Also, noise immunity of the proposed architecture is studied by injecting random fluctuations across the spatial locations in the visual field. Detailed transistor-level SPICE simulations show that the proposed structure can accurately and reliably detect complex motions of an object in a two-dimensional visual field.