Energy profile analysis of Zynq-7000 programmable SoC for embedded medical processing: Study on ECG arrhythmia detection

Electrocardiogram (ECG) analysis has been established as a key element regarding the evaluation of the human health status. The computational complexity along with the strict constraints of real-time assessment of a heart beat, has made the ECG analysis flow a very challenging application for embedded medical devices. Recent advancements in cyber-physical and IoT systems are transforming medical processing towards embedded and wearable devices, thus making energy consumption a first class design objective. In this work, we focus on analysing the power, performance and energy profiles of an ECG analysis and arrhythmia detection software pipeline during its execution on a ZYNQ-based SoC. We evaluate a large set of design alternatives spanning from a pure software-only implementation to HW/SW oriented designs, in which High-Level Synthesis capabilities are utilized. Using the medically validated MIT-BIH ECG database, we examine the efficiency and the sensitivity of the design solutions in different operating frequencies and examine three Quality of Service (QoS) levels concerning the sampling rate of the ECG signal.