Speaker Identification based on Hybrid Clustering and Radial Basis Function

Speaker identification is the computing task to identify an unknown identity based on the voice. A good speaker identification system must have a high accuracy rate to avoid invalid identity. Despite of last few decades’ efforts, accuracy rate in speaker identification is still low. In this paper, we propose a hybrid approach of unsupervised and supervised learning i.e. subtractive clustering and radial basis function(Sub-RBF).The proposed fused technique yields promising results because subtractive clustering is able to solve the initial guesses of cluster center and difficulty level to determine the number of cluster. Besides that, RBF has simple network structure and faster learning algorithm. In RBF input to output map uses the local approximations which will combine the linear approximations and causes the linear combinations with less weight. RBF neural network model uses subtractive clustering algorithm to select the hidden node centers for high training speed. In the meantime, the RBF network is trained with a regularization term so as to minimize the variances of the nodes in the hidden layer and to perform accurate prediction. Promising results are achieved to identify speaker using proposed fused approach. [Ann Y. T, Rahim M.S.M., Altameem A, Rehman A, Amin, I, M. Saba T. Speaker Identification based on Hybrid Clustering and Radial Basis Function. J Am Sci 2012;8(10):71-75]. (ISSN: 1545-1003). http://www.jofamericanscience.org. 12