Lightweight Polygonal Approximation-Based ECG Signal Processing Platform

Abstract

An electrocardiogram (ECG) signal is one of the most important bio-signals because it is caused by the heart's electrical activity. Therefore, ECG-signal analysis provides information about the heart's condition, especially heart disease. For ECG-signal processing, the original method applied to Internet-of-Things edge devices (such as wearable devices) is server-centric digital signal processing. Edge devices have some restrictions, such as small memory size, limited performance, and poor power supply. Although these devices perform only simple processing, including data acquisition and transmission, the devices' power consumption is high because of a large amount of communication. To solve this problem, we propose a polygonal approximation-based ECG-signal processing platform that is lightweight enough to be implemented in edge devices. In this platform, the ECG data are compressed to a small number of vertices by polygonal approximation, and only vertices are transmitted to the server. Therefore, the amount of communication decreases, thereby reducing the edge device's power consumption. The proposed platform was validated on a virtual edge device, which consists of a RaspberryPi 3 model B microcontroller (MCU) and HealthyPi v3. Results showed a 98% reduction in power consumption compared to a server-centric digital signal processor.