Singer Identification Based on Artificial Neural Network

Music is a vocal or instrumental sounds (or combined) presented in such a way as to create, harmony, and expression of emotion. Present days accurately singer identification is necessary for music indexing and retrieval purpose. This paper proposes a unique features extraction algorithm for singer identification. In this paper, seven singers with five vocal songs are considered for singer identification. The most potential Mel-Frequency-Cepstral Coefficient (MFCC) based feature extraction algorithm, and an artificial neural network (ANN) classifier has been applied for the singer identification purpose. Four multilayer neural network training algorithm such as Levenberg-Marquardt, Bayesian regularization Backpropagation, Scaled Conjugate Gradient, and One-step secant Backpropagation algorithm has been used to classify the seven different singers voices. Three different feature extraction technique, such as MFCC, MFCC with five different musical features and MFCC with ten different musical features has been considered for the feature extraction. The highest training and testing accuracy have been achieved through this algorithm is 98.3% and 88.6%. Classification accuracy varies with musical features and classification algorithms.