In the design of medical decision support systems for the diagnosis of certain diseases associated with heart valves, the segmentation process that is described as the identification of fundamental heart sounds, which are named Si and S2 in the literature, and the intervals systole and diastole, is an important stage. Not only the phonocardiography (PCG) records, which are recorded during examination, may include some external sounds that stem from speech and friction of stethoscope but also various physiological sounds such as wheezing, coughing and lung sounds. Similarly, murmurs, which are signs of particular heart diseases, are another factor that affects the performance of the segmentation process. In this study with the intent to overcome such difficulties, an algorithm that performs the segmentation task by processing an envelope signal obtained using discrete wavelet transform and Mel-frequency cepstral coefficients techniques jointly was developed. The performance of the algorithm is tested on both normal heart sounds, which do not include any sign of disease, and abnormal heart sounds. As a result of these tests, for normal heart sounds Si and S2 sounds were detected with 99, 49% recall and 94, 07% precision, while with 97, 53% recall and 90, 04% precision for abnormal heart sounds.