Ivan Magrin-Chagnolleau

A tutorial on text-independent speaker verification

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.

Application of time-frequency principal component analysis to text-independent speaker identification

We propose a formalism, called vector filtering of spectral trajectories, that allows the integration of a number of speech parameterization approaches (cepstral analysis, /spl Delta/ and /spl Delta//spl Delta/ parameterizations, auto-regressive vector modeling, ...) under a common formalism. We then propose a new filtering, called contextual principal components (CPC) or time-frequency principal components (TFPC). This filtering consists in extracting the principal components of the contextual covariance matrix, which is the covariance matrix of a sequence of vectors expanded by their context. We apply this new filtering in the framework of closed-set speaker identification, using a subset of the POLYCOST database. When using speaker-dependent TFPC filters, our results show a relative improvement of approximately 20% compared to the use of the classical cepstral coefficients augmented by their /spl Delta/-coefficients, which is significantly better with a 90% confidence level.

Experiments on speech tracking in audio documents using Gaussian mixture modeling

This paper deals with the tracking of speech segments in audio documents. We use a cepstral-based acoustic analysis and Gaussian mixture models for the representation of the training data. Three ways of scoring an audio document based on a frame-level likelihood calculation are proposed and compared. Our experiments are done on a database composed of television programs including news reports, advertisements, and documentaries. The best equal error rate obtained is approximately 12%.

Indexing telephone conversations by speakers using time-frequency principal component analysis

We present an algorithm for the tracking of target speakers in telephone conversations. Speaker tracking consists in retrieving, in an audio recording, segments which have been uttered by a target speaker. We also compare two speech analysis techniques. The first one is the time-frequency principal component analysis. It is a new speech analysis technique based on the extraction of the principal components of the contextual covariance matrix, which is the covariance matrix of feature vectors expanded by their time context. The other one is the classical cepstral analysis. Experiments are carried out on a subset of the switchboard database.

Application of vector filtering to pattern recognition

We present a new formalism, called vector filtering, which consists in transforming a sequence of vectors through a matrical filtering. This formalism allows one to unify a number of classical approaches. We also show how vector filtering can be integrated in a pattern recognition system. We then propose a new filtering, called contextual principal components, which consists in calculating principal components on vectors augmented by their context. Then, we apply the new filtering in the framework of text-independent speaker identification, which consists in identifying a speaker by the voice without knowledge about the phonetic content. By using this new filtering, we are able to decrease the identification error rate to roughly 20 % compared to a system using the classical cepstral coefficients augmented by their delta parameters.