| Summary: | 碩士 === 國立中興大學 === 通訊工程研究所 === 100 === Analysis of heart sound signals by auscultation helps diagnosing abnormal heart sounds effectively. The key to achieve that is to locate the normal heart sounds in a
cardiac cycle, from which a medical practitioner can predict the type of cardiac diseasein an early stage.
In this research, we propose an algorithm for automatic heart sound analysis which exploits the temporal
correlation property to detect heart sounds without
synchronic reference signals.
It firstly uses wavelet transform to reduce the amount of cardiac murmurs. Then,the temporal correlation property which results from the periodic pattern of heart sound is explored to help identify the locations the first heart sound(S1) and the second heart sound(S2), from which the existence of murmurs or extra heart sounds can be detected. Simulation results show that the correct detection rate is 87.81% for S1 and 87.46% for S2 when the heart sound is embedded with murmur. When there exists extra heart sounds, the proposed algorithm can achieve 100% detection of both
S1 and S2.
|
|---|
