Low complexity sensors for body area networks
Proceedings of the 3rd International Conference on PErvasive Technologies Related to Assistive Environments
Blood oxygen estimation from compressively sensed photoplethysmograph
WH '10 Wireless Health 2010
Why is context-aware computing less successful?
CASEMANS '11 Proceedings of the 5th ACM International Workshop on Context-Awareness for Self-Managing Systems
Resource-efficient and reliable long term wireless monitoring of the photoplethysmographic signal
Proceedings of the 2nd Conference on Wireless Health
Hi-index | 0.00 |
We describe an ultra low power pulse oximeter sensor for long term, non-invasive monitoring of SpO2 and heart rate in Body Area Networks (BAN). Commercial pulse oximeter sensors consume about 20-60 mW of power during continuous operation. Other researchers have shown that accurate and noise robust wireless pulse oximeter sensors can be designed to operate with as little as 1.5 mW. The LEDs consume bulk of the power budget in pulse oximeter sensors. In this work, we describe a compressed sensing approach to sample the photodetector output, so that the LEDs can be turned off for longer periods and thus save sensor power. We randomly sample Photoplethysmogram (PPG) signals with about 10-40x fewer samples than with uniform sampling and demonstrate that the accuracy of heart rate estimation and blood pressure estimation are not compromised, using MIMIC database. This provides power savings of the order of 10-40x for a pulse oximeter sensor, by reducing the duration LEDs need to be turned on.