It is estimated that between 15.6 and 19.6 million people are living with rheumatic heart disease (RHD) worldwide, accounting for about one million deaths annually and 60% of Africa's open heart surgeries. As RHD results in heart murmurs that are almost always audible during auscultation, a mobile phone-based automatic auscultation device has the potential to identify those individuals with a high risk of having RHD. Such a device would allow cost-effective treatment while not requiring expert training or expensive equipment. This paper addresses two of the major steps when processing heart sounds recordings using such a device: signal quality classification, which achieved over 90% accuracy, and heart sound segmentation, with 93.5% F₁ score. Future steps required to develop a fully automatic device are then discussed.