Rate-adaptive compressed-sensing and sparsity variance of biomedical signals

Abstract

Biomedical signals exhibit substantial variance in their sparsity, preventing conventional a-priori open-loop setting of the compressed sensing (CS) compression factor. In this work, we propose, analyze, and experimentally verify a rate-adaptive compressed-sensing system where the compression factor is modified automatically, based upon the sparsity of the input signal. Experimental results based on an embedded sensor platform exhibit a 16.2% improvement in power consumption for the proposed rate-adaptive CS versus traditional CS with a fixed compression factor. We also demonstrate the potential to improve this number to 24% through the use of an ultra low power processor in our embedded system.