Phonocardiographic signal analysis method using a modified hidden Markovmodel
Abstract Auscultation is an important diagnostic indicator for cardiovascularanalysis. Heart sound classification and analysis play an important role in the auscultativediagnosis. This study uses a combination of Mel-frequency cepstral coefficient (MFCC) and hiddenMarkov model (HMM) to efficiently extract the features for pre-processed heart sound cycles for thepurpose of classification. A system was developed for the interpretation of heart sounds acquired byphonocardiography using pattern recognition. The task of feature extraction was performed usingthree methods: time-domain feature, short-time Fourier transforms (STFT) and MFCC. The performancesof these feature extraction methods were then compared. The results demonstrated that the proposedmethod using MFCC yielded improved interpretative information. Following the feature extraction, anautomatic classification process was performed using HMM. Satisfactory classification results(sensitivity > or =0.952; specificity > or =0.953) were achieved fornormal subjects and those withvarious murmur characteristics. These results were based on 1398 datasets obtained from 41 recruitedsubjects and downloaded from a public domain. Constituents characteristics of heart sounds werealso evaluated using the proposed system. The findings herein suggest that the described system mayhave the potential to be used to assist doctors for a more objective diagnosis.
Keywords Heart Sounds; Markov Chains
Annals of Biomedical Engineering
0090-6964, Volume 35, Issue 3, 2007, Pages 3-374
Wang,P; Lim,CS; Chauhan,S; Foo,JY; Anantharaman,V
