This paper presents a low-cost portable signal acquisition and processing device with application to the medical diagnosis of heart diseases through the time and frequency analysis of the human heart sound. Human heart diseases leave identifiable traces in the heart sound that can be isolated using signal processing techniques. A hardware device consisting on a microphone-adapted stethoscope, an acquisition and amplification circuit and a dsPIC-based processing unit was developed, that can be successfully used for this purpose. An example algorithm, focused on heart sound murmurs, that can classify the heart sound in four health classes, was implemented. The algorithm considers the patterns observed from a database of healthy and non-healthy hearts. This exploratory prototype was tested in a group of hospital patients proving to be especially adequate to assist in the diagnostic of valvular heart diseases. This type of low-cost and simple-to-use device can be of great interest in developing countries, where universal medical care is not available, providing a cost-effective, non-specialised identification of cases in urgent need of medical attention.