The Concept of Neuromorphic Vision Systems based on Memristive Devices

Abstract

Here we propose the concept of neuromorphic analog memristive vision systems. The main feature of this concept is the rejection of analog-to-digital and digital-to-analog conversions when capturing input visual data for a spiking neural network (SNN) based on memristive devices. This can be achieved by combining photodiodes and memristors and directly feeding analog pulses from the output of such a circuit to the input of a SNN circuit. This concept relates to the field of in-memory and in-sensor computing and will makes it possible to create more compact, energy-efficient visual processing units for wearable, on-board and embedded electronics for such areas as robotics, the Internet of Things, neuroprosthetics and other practical applications in the field of artificial intelligence.