A Biosignal Platform for Validating Phonocardiogram-Based Heart Rate Estimation Using Electrocardiogram

Physiological signals contain essential information that is useful for healthcare. With the growth of wearable technology, physiological signals such as electrocardiograms (ECG) and phonocardiograms (PCG), can already be acquired at ease. Despite the advancements in hardware, signal processing platforms for multimodal physiological data have not seen the same rate of development. This project aims to address this gap by developing a biosignal platform that evaluates the performance of various peak detection algorithms for ECG and PCG signals. The biosignal platform integrates ECG and PCG acquisition and synchronization, signal processing, peak detection, and heart rate calculation into a single system. ECG signals were acquired using an OpenBCI Ganglion in a 3-lead ECG setup, while PCG signals were recorded using a Stemoscope digital stethoscope. Signal acquisition was synchronized to ensure that the signals were time-aligned. The signal processing and peak detection component was implemented by integrating open source libraries that contain machine learning and digital signal processing (DSP) models. The instantaneous heart rate (IHR) was calculated from successive R-peaks and S1 peaks for ECG and PCG respectively. The performance of the algorithms was evaluated through the mean absolute error of the IHR, comparing the ECG from the PCG estimates. A minimum MAE of 4.222 beats per minute (BPM) was achieved using the Hamilton algorithm for ECG and the DWT-based method for PCG. Aside from self-recorded data, the platform was also used to evaluate synchronous ECG and PCG data from the EPHNOGRAM data set.