A 0.8μm 32Mpixel Always-On CMOS Image Sensor with Windmill Pattern Edge Extraction and On-Chip DNN

Abstract

This paper presents a CMOS image sensor (CIS) that integrates two operation modes: a high-resolution viewing mode with 0.8μm 32Mpixels and a low-power always-on object recognition mode consuming 2.67 mW at 10 fps. The CIS features a unique windmill-pattern analog edge extraction circuit that is resilient to illumination variations. An on-chip DNN processor was implemented alongside a compact algorithm with only 12 KB for coefficients and 48 KB for working memory. The design incorporates separate circuit areas for high-speed viewing and low-power sensing modes, thereby ensuring optimal performance and energy efficiency.

View Publication