Feature extracted from wavelet eigenfunction estimation for text-independent speaker recognition

摘要

A new speaker feature extracted from wavelet eigenfunction estimation is described. The signal is decomposed through interpolating the scaling function. Wavelets can offer a significant computational advantage by reducing the dimensionality of the eigenvalue problem. Our results have shown that this wavelet feature introduced better performance than the other Karhunen-Loeve transfonn (KLT) with respect to the percentages of recognition. (C) 2003 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.