Heart diseases diagnosis using HMM

The bare ear and the stethoscope were until recently of great help in classifying most heart diseases especially those related to valve problems. The newly developed electronic stethoscope and phonocardiography represent useful tools for recording heart sound signals. In this paper a diagnostic technique for heart diseases using heart sounds is suggested. Wavelet decomposition and mel cepstrum are used for feature extraction. Classification of the different heart diseases is then done using hidden Markov models (HMM). Three different techniques have been used and compared. The obtained recognition rates (RR) were 97.3%, 98.2%, and 99.1%.