|
Descripción
|
|
|---|---|
| Through the present paper, the system developed by the GIAPSI research group for the 2010 NIST Speaker Recognition Evaluation (NIST SRE10) will be presented. The proposed system uses glottal estimate parameter templates and classical cepstral information to build a model for each speaker involved in the recognition process. ALIZE [1] open-source software has been used to create the GMM models for both background and target speakers. The performance of the system under different data conditions is presented, confirming the better performance of the system if compared to classical parameterization approaches. As this is the first approach of the group to a major international evaluation campaign, the main problems faced during the evaluation as well as future work and conclusions will be presented. | |
|
Internacional
|
No |
|
DOI
|
|
|
Edición del Libro
|
0 |
|
Editorial del Libro
|
|
|
ISBN
|
978-84-693-3389-1 |
|
Serie
|
|
|
Título del Libro
|
1er WTM-IP Workshop en Multibiométricas para la Identificación de Personas |
|
Desde página
|
40 |
|
Hasta página
|
46 |