Compensating mismatch in a dedicated pixel array for moving edge detection

This paper describes the implementation of a CMOS edge-detecting pixel array with on-chip adaptation that compensates for intrinsic variation between devices. The adaptation algorithm, based on asymmetric Hebbian training, uses spike timing to adapt out the effects of device mismatch and process variation. Each pixel adapts to background illuminance and provides high-pass filtering output with respect to local positive and negative illuminance transients. Sample chips have been manufactured using a 0.35 /spl mu/m CMOS technology, and the test results confirming the feasibility of the chosen approach are reported.