Please use this identifier to cite or link to this item: http://hdl.handle.net/10553/44104
Title: Characterization of a speaker from a telephone recording using speech quality measurements
Authors: Alonso, Jesús B. 
Travieso, Carlos M. 
De León, José
Ferrer, Miguel A. 
UNESCO Clasification: 3307 Tecnología electrónica
Keywords: Telephony , Speech , Pathology , Acoustic noise , Stability , Frequency , Hospitals , Cultural differences , Testing , Proposals
Issue Date: 2005
Publisher: 1071-6572
Journal: null 
Abstract: Starting from a telephone recording of a speaker, by means of subjective techniques based on the experience, it is possible to identify characteristics of the speaker like for example the linguistic origin, the cultural level, and the repetitive use of certain grammatical expressions. In this work, the preliminary results of applying techniques based on speech quality measurements for characterizing to a speaker are presented. This technique is objective and it is based on the behaviour of the fonator system. These techniques have been applied with success in the medical environment. In this paper, we propose the use speech features to improve the voice quality measurement, with the purpose of extracting characteristic of a speaker that allows the identification of the speaker. We have tested different measures based on voiced speech in the context of telephone recordings using healthy speakers (high voice quality) and with speakers which present laryngeal pathologies (low voice quality). The speakers used in the study are classified between healthy speakers or pathological speakers by mean an automatic classification with the purpose of justifying the effectiveness of the measures proposals.
URI: http://hdl.handle.net/10553/44104
ISBN: 0-7803-9245-0
ISSN: 1071-6572
Source: Proceedings - International Carnahan Conference on Security Technology[ISSN 1071-6572] (1594884)
Appears in Collections:Actas de Congresos

Show full item record

Page view(s)

16
checked on Nov 16, 2019

Google ScholarTM

Check

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.