No Detail Found!

Contribution to the vocal print recognition in arabic language.

Author(s)
Abdennaceur Kachouri Laboratoire d’ Electronique et des Technologies de I’ Information ‘LETT’, ENIS, Universite de Sfax-Bfax-B.P.W, Sfax, Tunisia
Tarak Hdiji Laboratoire d’ Electronique et des Technologies de I’ Information ‘LETT’, ENIS, Universite de Sfax-Bfax-B.P.W, Sfax, Tunisia
Zied Sakka Laboratoire d’ Electronique et des Technologies de I’ Information ‘LETT’, ENIS, Universite de Sfax-Bfax-B.P.W, Sfax, Tunisia
Mounir Samet Laboratoire d’ Electronique et des Technologies de I’ Information ‘LETT’, ENIS, Universite de Sfax-Bfax-B.P.W, Sfax, Tunisia
Abstract
The study presents a new database dedicated to speaker recognition applications. The main characteristics of this Arabic database are spoken by native and non-native speakers, a single session of sentence reading and relatively extensive speech samples suitable for learning person specific speech characteristics. This speech database is dedicated to the modeling and the representation of speakers. The representation consists in extracting parameters (MFCC: Mel Frequency Cepstral Coefficients or LPCC: Linear Prediction Cepstral Coefficients) that characterize the voice or a speaker`s vocal print from isolated words either linked from the Arabic database prepared for this work. The technique used in the phase of recognition adapted to this type of data and that showed more performance is the one of HMM (Hidden Markov Models). One tidy preparation of the training database and a good choice of entrance parameters permits to finish an effective model.
Publication Details
Page(s) 2560-2567
DOI DOI not available
Published Journal: Journal of Applied Sciences, Volume: 7, Issue: 18, Year: 2007

Request the full text from PASTIC

0

Citations

0

Downloads

38

Views