A Discrete Wavelet Packet Transform Based Approach for Selection of Artifact-Free Phonocardiogram Segment

摘要

Phonocardiogram (PCG) signals are often corrupted by different types of artifacts which make computer based heart sound signal analysis more challenging. It requires human intervention to select decent quality of signal subsequence that is free from artifacts. In this work, we have proposed a new technique of automatic artifact-free subsequence selection from PCG signal by the application of Discrete Wavelet Packet Transform (DWPT) and quasiperiodicity property of artifact-free heart sound signal. DWPT is used to identify those parts of the signal, where the artifacts are prominent. After this, quasi-periodicity property of PCG signal is used as quality measure to select artifact-free subsequences. The algorithm shows good results when tested on normal and five commonly occurring pathological heart sound signals.