Compression and Multi-Spectral Sensing for Video Based Physiological Monitoring

Abstract

Remote physiological monitoring is an active area of research that extends monitoring capabilities traditionally found in a clinical setting towards the home, telehealth, and beyond. In particular, there is interest in leveraging consumer electronic devices for sensing physiological characteristics such as heart rate, heart rate variability, and blood oxygen saturation. This thesis focuses on enhancing the understanding and usage of the sensing component for these applications to improve the performance and quality of cardio-physiological monitoring. First, a close relationship between the color spaces used for video compression and the color projection planes commonly used for heart rate estimation is identified. % that results in higher compression of the physiological signal. The study demonstrates the impact of this observation on real and synthetic data to provide a foundation to guide future video coding to optimize its configurations to better preserve the heart rate signal for health related applications. Second, an investigation with a commercial-off-the-shelf (COTS) multi-spectral sensor is presented with key observations related to the sampling rate, exposure settings, and multi-channel processing. These observations will enable better usage of the sensor for future studies and data collections that leverage the more precise spectral measurements from the multi-spectral sensor compared to standard RGB cameras.