On the use of dynamic information for speaker identification
Rosa María Ortega-Mendoza, Esaú Villatoro Tello, Luis Villaseñor-Pineda, Manuel Montes y Gómez, Eduardo F. Morales · Research in Computing Science · 2007
This paper describes a language independent method for speaker identification. This method is based on a novel characterization of the speech signal that captures the dynamic information contained in the cepstral coefficients. The proposed method was evaluated through several experiments on a corpus of Mexican speakers. The achieved results demonstrated the relevance of the signal characterization, reaching an identification accuracy as high as 97% under a multi-class scheme.