Analysis of Acoustic Cardiac Signals for Heart Rate Variability and Murmur Detection Using Nonnegative Matrix Factorization-Based Hierarchical Decomposition

The detection of heart rate variability (HRV) via cardiac auscultation examination can be a useful and inexpensive tool which, however, is challenging in the presence of pathological signals and murmurs. The aim of this research is to analyze acoustic cardiac signals for HRV and murmur detection. A novel method based on hierarchical decomposition of the single channel mixture using various nonnegative matrix factorization techniques is proposed, which provides unsupervised clustering of the underlying component signals. HRV is determined over the recovered normal cardiac acoustic signals. This novel decomposition technique is compared against the state-of-the-art techniques, experiments are performed using real-world clinical data, which show the potential significance of the proposed technique.