An Elementary Neuro-Morphic Circuit for Visual Motion Detection with Single-Electron Devices Based on Correlation Neural Networks

摘要

This paper proposes a bio-inspired single-electron circuit for detecting motion in projected images. Motion detection is a primary task performed in the retina as part of early vision processing. Based on motion detection models: the correlation model and spatiotemporal gradient model, a number of motion detecting circuits have been proposed and implemented with CMOS mediums. In this paper, based on the correlation model, we propose a possible single-electron circuit configuration that can detect motion in incident images, and demonstrate its basic performance with a one-dimensional construction. Through Monte-Carlo based computer simulations, we confirmed that this construction can compute motion in projected images.