Design of a spatiotemporal correlation filter for event-based sensors

摘要: This paper reports the design of a 1mW, 10ns-latency mixed signal system in 0.18μm CMOS which enables filtering out uncorrelated background activity event-based neuromorphic sensors. Background (BA) output dynamic vision sensors is caused by thermal noise and junction leakage current acting on switches connected to floating nodes pixels. The reported chip generates pass flag for spatiotemporally correlated events post-processing reduce communication/computation load improve information rate. A with 128×128 array 20×20μm2 cells has been designed. Each filter cell combines programmable spatial subsampling temporal window based integration. Power-gating used minimize power consumption only activating threshold detection communication circuits receiving an input event. correlation targets embedded visual auditory systems, where low average latency are critical.