An artificial retina with adaptive mechanisms and its application to retinal prosthesis

Abstract

The retina is a paradigm of novel image processing systems in which a real time computation with low power dissipation and a compact hardware is required. The basic structure underlying the computation of outer retinal neural network was elucidated with physiological experiments and computer simulations. We describe the spatio-temporal properties of outer retinal circuit with an equivalent resistive circuit model. Based on the model, an artificial retina was fabricated using analog CMOS VLSI circuit technology. The output of artificial retina emulates the center-surround antagonistic receptive field. The chip was incorporated with light-adaptive functions, which are inspired by physiological observations, to extract edges of a given image effectively. Preliminary experiments of applying the artificial retina for robot vision as well as for retinal prosthesis are performed.