Analysis of methods for heart sound cancellation in respiratory sounds

• Main Author: Smita Shekar.
• Other Authors: Farook Sattar
• Format: Theses and Dissertations
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation::Medical electronics
• Online Access: http://hdl.handle.net/10356/18786
• Institution: Nanyang Technological University

Respiratory sounds have since long been important indicators of a person’s health and disease. Advances in computer technology and ability of signal processing algorithms to detect symptoms and to derive characteristic features of the respiratory sounds for both diagnostic and assessment of treatment purposes has led to significant research in the field of respiratory sound analysis. In analysis of respiratory sounds, heart sounds form unavoidable interference during data recording and analysis. Thus the first step towards respiratory sound analysis is to detect and remove sections of heart sounds in respiratory sounds has been a topic of on-going research. Various methods have been formulated to achieve efficient localization and reconstruction of localized segments. In this report, in order to reduce the effect of heart sounds on respiratory sounds, we first localize heart sound segments. Two HS localization methods implemented are: heart sound localization using entropy of respiratory sounds and localization by a wavelet based approach using multiscale products. On observation, it is seen that heart sound localization by entropy approach shows better results. The results of both localization methods have been evaluated by comparing their False Positive and False Negative values. Comparison results on real recorded respiratory sounds how that the entropy based method is more effective. The localized HS include segments based on entropy approach are then used to reconstruct respiratory sound by linear prediction and by adaptive filtering using RLS algorithm. In case of Respiratory sound reconstruction, two methods based on linear prediction and adaptive filtering have been applied and the methods have been evaluated by comparing the Signal to Noise ratio before and after reconstruction. It is seen that linear prediction performs better when compared to Adaptive noise cancellation using RLS algorithm. The difference in performance and the underlying reasons have been discussed.