Practical Compression Methods for Quantized Compressed Sensing

In order to save energy of low-power sensors in Internet of Things applications, minimizing the number of bits to compress and communicate real-valued sources with a pre-defined distortion becomes crucial. In such a lossy source coding context, we study rate-distortion (RD) performance of various single-sensor quantized compressed sensing (QCS) schemes for compressing sparse signals via quantized/encoded noisy linear measurements. The paper combines and refines the recent advances of QCS algorithm designs and theoretical analysis. In particular, several practical symbol-by-symbol quantizer based QCS methods of different complexities relying on 1) compress-and-estimate, 2) estimate-and-compress, and 3) support-estimation-and-compress strategies are proposed. Simulation results demonstrate the RD performances of different schemes and compare them to the information-theoretic limits.