Understanding the Practical Limitations of Applying Analog Compressed Sensing Systems to ECG Signals

Abstract

Body area networks (BAN), networks of wearable and wireless physiological sensors, are expected to have a profound positive impact in healthcare. The bio-signal sensors are equipped with ultra-low power radios communicating to a BAN personal base station, and ultimately the healthcare provider. Most of the power dissipated in a state-of-the-art bio-signal sensor occurs when the RF power amplifier transmits data to the personal base station. Thus, a method is desired that decreases the amount of data to be transmitted which reduces the duty cycle of the power amplifier and increases the overall energy efficiency. Compressed sensing (CS) is a compression scheme capable of significantly reducing a signal acquisition's data rate. CS requires only a few incoherent measurements to compress signals that are sparse in some domain. Since compressed sensing is still an emerging topic, only a handful of CS systems have been realized in hardware. These systems have shown promising and yet limited abilities. The objective of this research is to provide designers with a roadmap that enables them to more easily make correct decisions in designing analog CS encoders and decoders for bio-signals. By showing the impact of the considerations of this CS system on ECG signals, it will set up a framework for how to approach and/or analyze the design of these systems for all bio-signals. The CS roadmap accomplishes this goal this by demonstrating the importance of signal sparsity, guides the design of sensing matrix generation, addressing the impact of several analog CS imperfections on CS compression and guides the selection of proper CS reconstruction algorithms.