An analog-digital hybrid CMOS circuit for two-dimensional motion detection with correlation neural networks

In this paper, we propose an analog-digital hybrid CMOS circuit producing two-dimensional optical flows, aiming at the development of high-speed and compact motion-sensing systems. The proposed circuit can compute the optical flow in real time by parallel operations of a number of local motion processors on the basis of the mechanisms of biological motion detectors. The circuit consists of asynchronous current-mode digital subcircuits for edge detection and analog subcircuits for local motion detection. The results of SPICE simulation confirmed that the circuits can compute two-dimensional local velocities with extracting edges of incident images. This shows that a high-speed motion-sensing chip can be developed with standard VLSI technology