Energy-Aware Design of Compressed Sensing Systems for Wireless Sensors Under Performance and Reliability Constraints

This paper describes the system design of a compressed sensing (CS) based source encoding system for data compression in wireless sensor applications. We examine the trade-off between the required transmission energy (compression performance) and desired recovered signal quality in the presence of practical non-idealities such as quantization noise, input signal noise and channel errors. The end-to-end system evaluation framework was designed to analyze CS performance under practical sensor settings. The evaluation shows that CS compression can enable over 10X in transmission energy savings while preserving the recovered signal quality to roughly 8 bits of precision. We further present low complexity error control schemes tailored to CS that further reduce the energy costs by 4X as well as diversity scheme to protect against burst errors. Results on a real electrocardiography (EKG) signal demonstrate 10X in energy reduction and corroborate the system analysis.