Forensic Application to Speaker Voice Recognition
Article Sidebar
Main Article Content
Abstract
In this article, we are interested in the Voice Recognition of Arabic Speakers with a view to Forensics, that is to say in the Criminalistics field (RVC) in particular on the two major tasks the Voice Identification of a Forensic Arabic Speaker (IVC) and Vocal Authentication of a Forensic Arabic Speaker (AVC) in text-independent mode. From the vocal signals of these speakers, information relating to their identities is extracted by cepstral analysis MFCC (Mel Frequency Cepstral Coefficients) with the latter by estimating GMM models (Gaussian Mixture Models) of robust speakers. Thus, a voice trace can be analyzed and subsequently compared with GMMs by applying the Bayesian approach, Likelihood-Ratio (LLR), in order to allow its identification and authentication. Our experiments carried out on the RVC system show that a GMM composed of 32 Gaussians is largely sufficient to represent the distribution of the vectors of a single speaker (the criminal) as well as the recording material which gives better performance of this elaborate system. . Indeed, we obtained satisfactory results. These can help the justice system to make a decision in order to resolve forensic problems
Article Details
References
61
62
Ouassila Kenai, Mhania Guerti
[6]
[8]
[9]
Science Casework”, Science & Justice, 38(3), 198-202, Suisse 1998. Andrzej D.. Meuwly D. and Alexander A., “Statistical Methods and Bayesian In- terpretation of Evidence in Forensic Au- tomatic Speaker Recognition”, Speech Processing Group, the Forensic Science Service. Gonzales-Rodriguez J, Ortega-Garcia J. and Sanchez-Bote J., “Forensic Identifi- cation Reporting Using Automatic Bio- metric Systems”, Speech and Signal Pro- cessing Group (ATVS), DIAC, Spain 2005. Kanungo T. and Mount D. M., “An Effi- cient k-Means Clustering Algorithm: Analysis and Implementation”, IEEE transactions on pattern analysis and ma- chine intelligence, Vol. 24, N° 7. July 2002. Dempster A. P., Laird N-M and Rubin D- B., “Maximum-likelihood from incom- plete data via the EM algorithm”, J. Roy- al Statist. Soc. Ser. B, 39, 1997. [10] Bilmes J., Gentle, A. “Tutorial of the EM Algorithm and its Application to Parame- ter Estimation for Gaussian Mixture and Hidden Markov Models”, Report, Uni- versity of Berkeley, ICSI-TR-97-021, 1997. [11] Burget L., Matejka P., Schwarz P. Glembek O. and Cemocky J-H., “Analy- sis of Feature Extraction and Channel Compensation in a GMM Speaker Recognition System”, IEEE Transactions on Audio, Speech, and Language Pro- cessing, Vol 15, N° 7, 2007. [12] Amrous A-L, Debyeche M. and Krobba À.… “Coopération de connaissances dans les modéles de Markov cachés pour la re- connaissance de mots isolés arabes”, 1st International Conference on Image and Signal Processing and their Applications, Mostaganem, Algérie 2009. [13] Kenai O. and Guerti M., “Identification d’un Locuteur Arabophone en vue la Criminalistique”, | Third — International Conference on Industrial Engineering & Manufacturing, Université de Batna, 4-5,
2013.