Cardiac auscultation is the examination of the heart by listening to the sound produced by the heart through a stethoscope. Heart sounds can provide information about the functioning of the heart valve condition as well as information about the structural abnormalities of the heart. However, it needs intensive training for mastering. The objective of the work was to develop an algorithm for heart sound signal classification applied to computer-assisted digital auscultation in order to identify pathological events. The method can be described as follow: data collection, pre-processing, segmentation, feature extraction, and classification. The data were collected from volunteers using an electronic stethoscope and from a database available in the internet. The heart sound data were then extracted and split into training, validation, and testing datasets. In the training process the dataset was labeled as normal and abnormal (aortic stenosis, mitral stenosis, aortic regurgitation, pulmonic regurgitation, tricuspid stenosis, flow murmur, and patent ductus arteriosus). The convolution neural network is used as a classifier in the learning process to obtain the learning model. The model was validated and tested using the available datasets. The experimental results show that the algorithm has the capability to classify the heart sound into normal and abnormal with a high detection rate.