Sparse Representation Classifier (SRC) is one of the popular and efficient methods of classification for biometric traits. Redundant dictionaries for training samples are created in SRC classification which requires complex mathematical computation. This complexity further increases when the entire training set of samples is used for the classification of the test sample. In this paper, an efficient heart sound recognition algorithm is proposed based on a combination of the two classification methods, namely, Nearest Centroid Neighbor and Sparse Representation Classification (kNCN-SRC). In this method, firstly, k nearest centroid neighbors for the test sample are computed, and then the test sample is classified by sparse representation classification with respect to the k selected nearest neighbors. The proposed kNCN-SRC method showed a significantly increased recognition rate of 8.63% when compared to that of the SRC. This improved recognition rate is due to the selection of nearest neighbors as training signals for classification by SRC. Also, as the selection of training signals is based on the nearest centroid neighbor, this improves the recognition rate as the best training signals are selected for classification by SRC. The findings of the present study showed that the kNCN-SRC classification method demonstrated an improved recognition rate and was found to be a more suitable classifier than SRC for heart sound biometric systems