Using Dysphonic Voice to Characterize Speaker’s Biometry

Phonation distortion leaves relevant marks in a speaker’s biometricpro le. Dysphonic voice production may be used for biometrical speaker charac-terization. In the present paper phonation features derived from the glottal source(GS) parameterization, after vocal tract inversion, is proposed for dysp...

Descripción completa

Detalles Bibliográficos
Autores: Gómez, Pedro, San Segundo Fernández, Eugenia, Mazaira, Luis M., Álvarez, Agustín, Rodellar, Victoria
Tipo de recurso: artículo
Estado:Versión publicada
Fecha de publicación:2014
País:España
Institución:Consejo Superior de Investigaciones Científicas (CSIC)
Repositorio:DIGITAL.CSIC. Repositorio Institucional del CSIC
OAI Identifier:oai:digital.csic.es:10261/362907
Acceso en línea:http://hdl.handle.net/10261/362907
Access Level:acceso abierto
Palabra clave:Phonation
Speaker Recognition
Voice Production
Speech Processing
Linguistic research
Descripción
Sumario:Phonation distortion leaves relevant marks in a speaker’s biometricpro le. Dysphonic voice production may be used for biometrical speaker charac-terization. In the present paper phonation features derived from the glottal source(GS) parameterization, after vocal tract inversion, is proposed for dysphonic voicecharacterization in Speaker Veri cation tasks. The glottal source derived param-eters are matched in a forensic evaluation framework de ning a distance-basedmetric speci cation. The phonation segments used in the study are derived from llers, long vowels, and other phonation segments produced in spontaneous tele-phone conversations. Phonated segments from a telephonic database of 100 maleSpanish native speakers are combined in a 10-fold cross-validation task to producethe set of quality measurements outlined in the paper. Shimmer, mucosal wavecorrelate, vocal fold cover biomechanical parameter unbalance and a subset of theGS cepstral pro le produce accuracy rates as high as 99.57 for a wide threshold in-terval (62.08-75.04%). An Equal Error Rate of 0.64 % can be granted. The proposedmetric framework is shown to behave more fairly than classical likelihood ratiosin supporting the hypothesis of the defense vs that of the prosecution, thus offeringa more reliable evaluation scoring. Possible applications are Speaker Veri cationand Dysphonic Voice Grading