Jean-François Bonastre
Université d'Avignon
Session JEP poster P2 Lundi 9 Juin - 16h00 18h00
-
papier 1644
Surveillance vocale de réseaux de communication professionnels par la reconnaissance du locuteur
- Alexandre Preti ( Thales Commmunications / LIA)
- Bertrand Ravera ( Thales Commmunications)
- François Capman ( Thales Commmunications)
- Jean-François Bonastre ( Université d'Avignon / Laboratoire d'Informatique d'Avignon)
- Résumé : Even if the speaker recognition field is very dynamic, few studies concern the constraints linked to the use of a speaker recognition system inside a professional telecommunication network. This paper deals with this problem and proposes some adaptation of such system in the focus of a real world network monitoring application. Both real-time constraints and distributed architectures are investigated. We propose a frame-by-frame on-line processing for feature extraction, frame selection and normalization. The links between the network speech coder and the speaker recognition system are also investigated, for both the ETSI TETRA speech codec (at 4600 bit/sec) and the NATO STANAG 4591 (at 2400 bit/sec). The proposed solutions are compared with a classical unconstrained front-end (off-line processing).
- article
Session JEP poster P3 Mardi 10 Juin - 14h00 16h00
-
papier 1628
Utilisation de la structure de mots de passe personnalisés pour la reconnaissance de locuteurs embarquée
- Anthony Larcher ( Université d'Avignon, LIA)
- Jean-François Bonastre ( Université d'Avignon, LIA)
- John-S.d. Mason ( Speech and Image group, Swansea University)
- Résumé : Embedded speaker recognition in mobile devices involves a limited amount of computing resource. Both the enrolment and the test have to be done using short audio sequences. Even if they proved their efficiency in more classical situations, GMM/UBM based systems show their limits in this context. This paper deals with this problem and proposes to take into account the linguistic nature of the speech material inside the GMM/UBM framework. The proposed solution mixes the text-independent aspects of the GMM/UBM with a semi-continuous like approach in order to deal with the text-dependent information. This system respects both the resource and the ergonomic constraints of the considered application field. The preliminary experiments are done on the MyIdea database and show the potential of the proposed approach.
- article
Session JEP poster P3 Mardi 10 Juin - 14h00 16h00
-
papier 1657
Adaptation rapide de modèles acoustiques compacts
- Christophe Lévy ( Université d'Avignon et des Pays de Vaucluse)
- Georges Linarès ( Université d'Avignon et des Pays de Vaucluse)
- Jean-François Bonastre ( Université d'Avignon et des Pays de Vaucluse)
- Résumé : In a previous work we presented a new architecture dedicated to embedded speech recognition. It relies on a general GMM, which represents the whole acoustic space, associated with a set of HMM state-dependent probability functions modeled as transformations of this GMM. This work takes advantage of this architecture to propose a fast and efficient way to adapt the acoustic models. The adaptation is performed only on the general GMM model and does not require state-dependent adaptation data. It is also very efficient in terms of computational cost. We evaluate our approach in the voice-command task. This adaptation method achieved a relative error-rate decrease of about 10% even if few adaptation data are available.
- article
Session JEP poster P4 Mardi 10 Juin - 14h00 16h00
-
papier 1669
Analyse des scores imposteurs d'un Système de VAL GMM-UBM
- Salah-Eddine Mezaache ( Laboratoire d'Informatique d'Avignon (LIA))
- Driss Matrouf ( Laboratoire d'Informatique d'Avignon (LIA))
- Jean-François Bonastre ( Laboratoire d'Informatique d'Avignon (LIA))
- Résumé : In this paper, we present an analysis of the problem of impostors trials with high scores in the context of NISSRE- 2006 eval [4]. Trials are based on LIA-GMMUBM refernce system [5]. We propose a method to deal with such trials called REVERSE method. Thus, less than a 1% trials on the NIST-2006 raise the DCFmin of 40%. Our motivation were to perform investigation on impostor scores, attempt to understand
- article
Session JEP orale O2 Pathologies Mardi 10 Juin - 16h30 18h30
-
papier 1629
Analyse Phonétique dans le Domaine Fréquentiel pour la Classification des Voix Dysphoniques
- Gilles Pouchoulin ( Université d'Avignon et des Pays de Vaucluse, Laboratoire Informatique d'Avignon (LIA))
- Corinne Fredouille ( Université d'Avignon et des Pays de Vaucluse, Laboratoire Informatique d'Avignon (LIA))
- Jean-François Bonastre ( Université d'Avignon et des Pays de Vaucluse, Laboratoire Informatique d'Avignon (LIA))
- Alain Ghio ( Université de Provence, Laboratoire Parole et Langage (CNRS-LPL))
- Antoine Giovanni ( Université de Provence, Laboratoire Parole et Langage (CNRS-LPL))
- Résumé : Concerned with pathological voice assessment, this paper aims at characterizing dysphonia in the frequency domain for a better understanding of related phenomena while most of the studies have focused only on improving classification systems for diagnosis help purposes. Based on a first study which demonstrates that the low freqencies ([0-3000]Hz) are more relevant for dysphonia discrimination compared with higher frequencies, the authors propose in this paper to pursue by analyzing the impact of the restricted frequency subband ([0-3000]Hz) on the dysphonic voice discrimination from a phonetical point of view. In this sense, performance of the GMM-based automatic dysphonic voice classification system is measured according to different phoneme classes and frequency bands ([0-3000] and [0-8000]Hz).
- article
Session JEP orale O4 Reconnaissance de la parole et du locuteur Jeudi 12 Juin - 14h00 16h00
-
papier 1626
La reconnaissance du locuteur : un problème résolu ?
- Jean-François Bonastre ( Université d'Avignon)
- Driss Matrouf ( Université d'Avignon)
- Résumé : Cet article présente un court résumé des progrès réalisés ces dernières années en Reconnaissance du Locuteur. Il tente de montrer qu'en dépit de l'impressionant gain enregistré en termes de réduction des taux d'erreurs, plusieurs questions restent ouvertes. Le papier conclut en ouvrant une série de pistes de recherche pour la reconnaissance du locuteur.
- article