A Pixelated Monolithic CMOS PPG Sensor for Spatial Feature Acquisition

Abstract

This article presents a pixelated monolithic CMOS photoplethysmography (PPG) sensor. The spiking neural network (SNN)-inspired architecture, which includes an unbiased photodiode (PD)-based light-to-digital converter (LDC), efficiently converts spatial features of PPG signals to multiple pixelated outputs. The sensor consists of 144 light-sensing neurons (LSNs) and simultaneously generates 1 b 12 row/12 column-wise pixelated outputs. The spatial features of the PPG signals are thoroughly studied with measurement results. Moreover, ambient light canceling with independent component analysis (ICA) is demonstrated as a key application of the proposed sensor. The sensor chip achieves 114 dB dynamic range while consuming 23.3 μW at 1-V supply voltage.