Cardiac auscultation that is a commonly used method to diagnose heart murmurs caused by cardiac disorders. Taking into account that this method is quite subjective and time consuming, the computerization of diagnosis process would significantly enhance clinical auscultation. Development of automated auscultative diagnosis systems, which provide more objective and reliable results, would be beneficial to reduce the classification errors for the cardiac disorders. The presented study uses a combination of Mel-frequency cepstral coefficient (MFCC), BaumWelch parameter re-estimation and Hidden Markov Model (HMM) to diagnose and categorize heart murmurs. Classification experiments were conducted on the 84 high-quality heart sound data made up of 6 different types of murmurs. From this, average correct classification rate of 98.8% was achieved when the HMM has 5 states and frame size is 25ms. This study shows that, a highly successful automated auscultative diagnosis system working on less feature can be developed as a supportive diagnostic tool for health-care professionals.