Heart sound classification using hidden markov model

Abstract

Cardiovascular disease (CVD) is among the leading life threatening ailments [1] [2] .Under normal circumstances, a cardiac examination utilizing electrocardiogram appliances or tools is proposed for a person stricken with a heart disorder, The logging of irregular heart behaviour and morphology is frequently achieved through an electrocardiogram (ECG) produced by an electrocardiographic appliance for tracing cardiac activity, For the most part, gauging of this activity is achieved through a non-invasive procedure l.e. through skin electrodes, Taking into consideration the ECG and heart sound together with clinical indications, the cardiologist arrives at a diagnosis on the condition of the patient's heart, This paper focuses on the concerns stated above and utilizes the signal processing theory to pave the way for better heart auscultation performance by GPs, The objective is to take note of heart sounds in correspondence to the valves as these sounds are a source of critical Information.Comparative investigations regarding MFCC features with varying numbers of HMM states and varying numbers of Gaussian mixtures were carried out for the purpose of determining the impact of these features on the classification implementation at the sites of heart sound auscultation.