An implementation of magnocellular pathways in event-based retinomorphic systems

In this paper we present an event-based system for emulating the functional behaviour of retinal ganglion cells (RGCs) with transient non-linear responses and large receptive fields. The system uses a Dynamic Vision Sensor (DVS), which generates asynchronous events in response to reflectance changes, as input to a microcontroller processing unit. The processing unit then implements an interrupt driven model of an Approach Sensitive RGC (AS-RGC) and displays the result on an 8x8 LED array. We describe how, for this type of input, a simplified Integrate-and-Fire neuron model can be derived that uses the asynchronous events to represent changes in the post-synaptic potentials of the AS-RGC. The system's functional behaviour is demonstrated using two pertinent examples: an expanding black square representing approach motion; and a moving black bar representing lateral motion.