Comparison of speaker recognition systems on a real forensic benchmark
نویسندگان
چکیده
This paper analyses the performance of several automatic speaker recognition systems using a real forensic database. The systems evaluated have been tested or are currently in use by forensic institutes. A comprehensive error analysis is performed in order to assess the each system’s behaviour to real casework. We further investigate compensation techniques aimed at minimising the performance gap between laboratory development and application on real forensic data. While unrestricted application of automatic systems in the forensic domain is still not a reality, our experiments suggest that automatic systems can be a valuable support in decisionmaking for the forensic examiner.
منابع مشابه
Statistical methods and Bayesian interpretation of evidence in forensic automatic speaker recognition
The goal of this paper is to establish a robust methodology for forensic automatic speaker recognition (FASR) based on sound statistical and probabilistic methods, and validated using databases recorded in real-life conditions. The interpretation of recorded speech as evidence in the forensic context presents particular challenges. The means proposed for dealing with them is through Bayesian in...
متن کاملAutomatic forensic voice comparison (Automatischer forensischer Stimmenvergleich)
Automatic speaker recognition systems have been developed and constantly improved during recent decades. However, even though these systems have been applied to the task of forensic voice comparison since the early 2000s, their behaviour under realistic forensic conditions has not often been thoroughly tested since then. Additionally, the application of these systems is not standardised and is ...
متن کاملAn information theory based data-homogeneity measure for voice comparison
In forensic voice comparison, it is strongly recommended to follow the Bayesian paradigm to present a forensic evidence to the court. In this paradigm, the strength of the forensic evidence is summarized by a likelihood ratio (LR). Theoretically, a LR embeds intrinsically the reliability information. So the LR could belong to large values in good conditions, about 10±10, while in bad conditions...
متن کاملOn the Application of the Bayesian Framework to Real Forensic Conditions with GMM-based Systems
In this paper, excellent results are provided in the calculation of likelihood-ratios in real forensic conditions within the bayesian framework for the evaluation of speech evidences with a GMM-based speaker recognition system. Reported experiments have been performed with speakers from the Ahumada/Gaudí database, where 249 (122 male and 127 female) acted as reference population for the evaluat...
متن کاملForensic Speaker Verification in Noisy Environmental by Enhancing the Speech Signal Using ICA Approach
We propose a system to real environmental noise and channel mismatch for forensic speaker verification systems. This method is based on suppressing various types of real environmental noise by using independent component analysis (ICA) algorithm. The enhanced speech signal is applied to mel frequency cepstral coefficients (MFCC) or MFCC feature warping to extract the essential characteristics o...
متن کامل