The phonocardiogram (PCG) is an easy and costless yet powerful tool to detect the heart condition. While the cardiovascular disease is an increasing global threat to humanity, there is a decrease in doctors' capability for diagnosing these diseases by auscultation. In this work, a neural network based automated system is proposed to aid early detection of these diseases. Two template-based feature representations are developed to effectively represent the characteristics of heart sound. These features, extracted from sound files of known cases, are then used to train a neural network. Our experimental results corroborate that the proposed method can efficiently detect the heart condition with good overall classification accuracy. In addition, a graphical user interface and a low cost device is developed which is user-friendly and can be used without much relevant knowledge.