This study implements a digital stethoscope by integrating analogy stethoscope and computer, by which physicians can not only perform diagnosis via traditional analogy stethoscope, but also they can analyze the data stored in the computer due to the advantages of digitization. To escape from the impact of man-made noise in the digital stethoscope, an algorithm for extracting effective heart sound signal is developed by using wavelet de-noising, autocorrelation function, and probability statistics. Physicians can utilize the extracted heart sound signal to execute prompt diagnosis. Meanwhile, a scheme for detecting abnormal heart sound is proposed to classify heart murmur. Experimental results verify potential of the proposed system in practical diagnosis application.