Blood oxygen estimation from compressively sensed photoplethysmograph

In this work, we consider low power, wearable pulse oximeter sensors for ambulatory, remote vital signs monitoring applications. It is extremely important for such sensors to maintain clinical accuracy and yet provide power savings to enable non-intrusive, long lasting sensors. Our contributions in this work include sub-Nyquist, random sampling of evanescent red and infra red (IR) photoplethysmograph (PPG) signals in real time under the Compressed Sensing (CS) paradigm. We describe the real time platform and demonstrate that the SpO2 accuracy is not compromised due to aliasing of ambient light artifacts, even when average number of measurements is much below that of Nyquist rate. We briefly discuss the various modules contributing to overall power consumption of a wireless pulse oximeter sensor and show that 10x reductions in LED power and radio power are possible, without sacrificing the SpO2 accuracy.