Robust heart rate estimation from noisy phonocardiograms

Accurate heart rate estimation is a fundamental process when analysing phonocardiograms (PCGs). While this is trivial in noise-free recordings, it becomes a difficult task in PCGs corrupted by various noise sources. While numerous PCG-based heart rate estimation techniques have been explored in the literature, no comparison between these techniques has been performed to identify the best-performing method in noisy recordings. This paper evaluates various denoising, normalisation, envelope extraction and heart rate estimation techniques on 585 noisy recordings made using four different devices. The best-performing algorithm correctly estimated the heart rate in 471 (80.5%) of these PCGs, while correctly estimating the heart rate in 86% of the PCGS from a separate (publicly available) test dataset.