A Tutorial on Text-Independent Speaker Verification

Abstract

<p/> <p>This paper presents an overview of a state-of-the-art text-independent speaker verification system. First, an introduction proposes a modular scheme of the training and test phases of a speaker verification system. Then, the most commonly speech parameterization used in speaker verification, namely, cepstral analysis, is detailed. Gaussian mixture modeling, which is the speaker modeling technique used in most systems, is then explained. A few speaker modeling alternatives, namely, neural networks and support vector machines, are mentioned. Normalization of scores is then explained, as this is a very important step to deal with real-world data. The evaluation of a speaker verification system is then detailed, and the detection error trade-off (DET) curve is explained. Several extensions of speaker verification are then enumerated, including speaker tracking and segmentation by speakers. Then, some applications of speaker verification are proposed, including on-site applications, remote applications, applications relative to structuring audio information, and games. Issues concerning the forensic area are then recalled, as we believe it is very important to inform people about the actual performance and limitations of speaker verification systems. This paper concludes by giving a few research trends in speaker verification for the next couple of years.</p

Effects of Long-Term Ageing on Speaker Verification

“…During testing, the likelihood of the test data given both a speaker's GMM and the UBM were calculated. Scoring was then done using the standard likelihood ratio framework [14], by subtracting the log likelihood score of the UBM from the log likelihood score of the speaker model.…”

Score normalization for keystroke dynamics biometrics

“…The rationale behind this statement is that score normalization reduces the misalignment between score distributions from different users. There are several works in the literature [5][6][7][8][9][10][11][12] analyzing the effectiveness of score normalization techniques in behavioral biometrics (e.g. voice and signature).…”

“…Score normalization has proved its usefulness for improving the performance of behavioral biometric traits such as signature [5][6][7][8] or voice [9][10][11][12]. The normalization of score mitigates the effects of misalignment between scores distribution from different users (this misalignment is common in behavioral traits).…”

The IIR Submission to CSLP 2006 Speaker Recognition Evaluation

“…Table 1) out of the six evaluation categories (see shaded boxes in Table 1) of this year SRE event. Our submission is built upon four subsystems using speaker information from acoustic spectral features [2,5,6,7], as illustrated in Fig. 1.…”

“…1. The speaker information represented in various forms is modeled using Gaussian mixture model (GMM) [7,8] and support vector machine (SVM) [9,10]. Feature extraction and speaker modeling techniques employed in the subsystems are described in Section 2 and Section 3, respectively.…”