Abstract
Retinomorphic chips may improve their spike-coding efficiency by emulating the primate retina's parallel pathways. To this end, I recreated retinal microcircuits in a chip, Visio1, that models the four predominant ganglion-cell types. It has 104/spl times/96 photoreceptors, 4/spl times/52/spl times/48 ganglion-cells, a die size of 9.25/spl times/9.67 mm/sup 2/ in 1.2 /spl mu/m 5V CMOS, and consumes 11.5 mW at 5 spikes/second/neuron. Visio1 includes novel subthreshold current-mode circuits that use horizontal-cell autofeedback to decouple spatiotemporal bandpass filtering from local gain control and use amacrine-cell loop-gain modulation to adapt highpass and lowpass temporal filtering. Different ganglion cells respond to motion in a stereotyped sequence, making it possible to detect edges of one contrast or the other moving in one direction or the other. The author presents results from a multichip 2-D motion architecture, which implements Watson and Ahumada's model of human visual-motion sensing.

This publication has 28 references indexed in Scilit: