期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2014
卷号：65
期号：3
出版社：Journal of Theoretical and Applied
摘要：In this paper, we propose a speaker recognition system based on features extracted from the speech recorded using close speaking microphone in clean and noisy environment. This system recognizes the speakers from a number of acoustic features that include linear predictive coefficients (LPC), linear predictive cepstral coefficients (LPCC) and Mel-frequency cepstral coefficients (MFCC). RBFNN and AANN are two modelling techniques used to capture the features. RBFNN model enables nonlinear transformation followed by linear transformation to achieve a higher dimension in the hidden space. The proposed work compares the performance of RBFNN with Autoassociative neural network (AANN). The autoassociative neural network (AANN) is used to capture the distribution of the acoustic feature vectors in the feature space. This model captures the distribution of the acoustic features of a class, and the backpropagation learning algorithm is used to adjust the weights of the network to minimize the mean square error for each feature vector. The experimental results show that, the performance of AANN performs better than RBFNN. AANN gives an accuracy of 94.93% for various acoustic features both in clean and noisy environment.
关键词：Radial Basis Function Neural Network (RBFNN); Autoassociative Neural Network (AANN); Linear Predictive Coefficients (LPC); Linear Predictive Cepstral Coefficients (LPCC); Mel-Frequency Cepstral Coefficients (MFCC); Speaker Recognition (SR)
