Application of AMDF for Vital Sensing and Its Implementation Toward IoT Edge Computing

Abstract

Non-contact vital sensing has increased its importance associated with the Internet of Things (IoT) concept. Although it creates opportunities to conceive new services, data collection using narrow-band wireless links is crucial. Edge computing architecture enables us to solve this problem even though it increases the power consumption in sensor nodes. Successful implementation of edge computing in the IoT system requires a power-efficient implementation of algorithms for data analysis and hardware configuration. This paper proposes to modify the average magnitude difference function (AMDF), which has been employed in pitch extractions of speech signals, and to apply it to the estimation of the respiration period measured by a Doppler sensor. In addition, a multi-stage search algorithm is proposed to be associated with an efficient hardware configuration. The performance of the algorithm and the implementation using a small-scale processor is verified through computer simulations and experiments.

Keywords

• Doppler sensing
• vital signs
• AMDF
• edge computing
• parameter estimation