A comparative study of phonocardiogram analysis techniques based on mixed signal processor

Normal heart sounds provide an indication of the general state of the heart in terms of rhythm and contractility. Any cardiovascular abnormality introduces additional sounds, and murmurs that changes the temporal nature of the acoustic signature. Such variations provide vital information in diagnosis of cardiac disorders. However, at the time of signal acquisition, the heart sound gets mixed with lung sound in both spatial and temporal domain which cannot be separated through linear filtering technique as the frequency ranges of both the signals are overlapping. To overcome this problem, we comparatively study three parametric signal processing techniques viz. Auto Regressive (AR), Auto Regressive Moving Average (ARMA) and Adaptive Line Enhancer (ALE) using Least Mean Square (LMS). We also compare their performance with wavelet based denoising strategies. The algorithms have been implemented and their performance is evaluated on the basis of Power Spectral Density (PSD) and percentage of heart sound recovery. The scope of the study also incorporated the applicability of the techniques for implementation on embedded mixed signal processing targets.