A Robust Heart Rate Detection Pipeline of Wearable ECG Signals in Motion

Wearable electrocardiogram (ECG) devices are at the forefront of wearable technology and are used for health monitoring, exercise, and physical training. Analyzing the data from these devices is challenging because of the high levels of noise and the presence of motion artifacts, which are less prevalent in nonwearable devices. Much research has been conducted on analyzing wearable ECG signals, but few studies provide robust algorithms that remove the very high levels of noise introduced specifically when subjects perform physical activities. In this paper, we introduce a novel and robust pipeline, which uses lowpass digital filtering, stationary wavelet transform, instantaneous amplitude, peak-finding methods, and Hampel filtering to de-noise ECG signals and provide accurate measurements of subjects' heart rates obtained from a three-electrode ECG armband.