Quasi-periodic modeling for heart sound localization and suppression in lung sounds

We present a novel approach to treating lung sounds contaminated by heart sounds by means of quasi-periodic signal modeling. Heart sounds are described through the long and short-term quasi-periodicity generated by cardiac cycle repetition and oscillations caused by valves openings and closures, respectively. In terms of signals, these quasi-periodicities drive time-variant HS envelope and phase which are modeled by single and piecewise time polynomials, respectively. Single polynomials account for slow and continuous envelope time variations, while piecewise polynomials capture fast and abrupt phase changes in short time intervals. Such a compact signal description provides an efficient way to fundamental heart sound (FHS) components localization and posterior removal from lung sounds. The results show that the proposed method outperforms two reference methods for medium (15 ml/s/kg) and high (22.5 ml/s/kg) air flow rates.