A hybrid parameterization technique for Speaker Identification
Pedro Gómez‐Vilda, Agustín Álvarez-Marquina, L. M. Mazaira, Rubén Fernández Pozo, Victor M. Garca Nieto, R. Martínez, Cristina Muñoz, V. Rodellar · UPM Digital Archive (Technical University of Madrid) · 2008
Classical parameterization techniques for Speaker Identifi-cation use the codification of the power spectral density of raw speech, not discriminating between articulatory features produced by vocal tract dynamics (acoustic-phonetics) from glottal source biometry. Through the present paper a study is conducted to separate voicing fragments of speech into vo-cal and glottal components, dominated respectively by the vocal tract transfer function estimated adaptively to track the acoustic-phonetic sequence of the message, and by the glottal characteristics of the speaker and the phonation ges-ture. The separation methodology is based in Joint Process Estimation under the uncorrelation hypothesis between vo-cal and glottal spectral distributions. Its application on voiced speech is presented in the time and frequency do-mains. The parameterization methodology is also described. Speaker Identification experiments conducted on 245 speak-ers are shown comparing different parameterization strate-gies. The results confirm the better performance of de-coupled parameterization compared against approaches based on plain speech parameterization. 1.