Speaker Recognition on Single- and Multispeaker Data

Weber, Frederick; Peskin, Barbara; Newman, Michael; Corrada-Emmanuel, Andrés; Gillick, Larry

doi:10.1006/dspr.1999.0362

Cited by 18 publications

(10 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is clear that the additional information available from the LVCSR model significantly improves speaker ID performance in the Extended Data Task environment. This is contrasted with earlier GMM and LVCSR results on 3 and 30 second tests from the NIST 1998 Speaker ID Evaluation [2], where the LVCSR system lags behind the GMM approach until at least 30 seconds of test data are available for analysis. …”

Section: Baseline Systems: Gmm and Lvcsrmentioning

confidence: 62%

“…Dragon Systems' GMM and LVCSR-based speaker ID systems are described in detail in [2]. We provide here an outline of their operation and performance on the Extended Data Task.…”

Section: Baseline Systems: Gmm and Lvcsrmentioning

confidence: 99%

“…However, as the amount of test and training data increases it becomes attractive to make use of speaker-specific characteristics which involve larger time scales, such as prosodic patterns. In this paper we will describe two standard speaker ID systems: a Gaussian mixture model (GMM) approach and a system based on large vocabulary continuous speech recognition (LVCSR), which were developed as part of Dragon Systems' regular participation in the annual NIST Speaker ID evaluations [2]. We then examine how larger time-scale information can be added to enhance the performance of these baseline systems.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Using prosodic and lexical information for speaker identification

Weber¹,

Manganaro²,

Peskin

et al. 2002

IEEE International Conference on Acoustics Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

We investigate the incorporation of larger time-scale information, such as prosody, into standard speaker ID systems. Our study is based on the Extended Data Task of the NIST 2001 Speaker ID evaluation, which provides much more test and training data than has traditionally been available to similar speaker ID investigations. In addition, we have had access to a detailed prosodic feature database of Switchboard-I conversations, including data not previously applied to speaker ID. We describe two baseline acoustic systems, an approach using Gaussian Mixture Models, and an LVCSR-based speaker ID system. These results are compared to and combined with two larger time-scale systems: a system based on an "idiolect" language model, and a system making use of the contents of the prosody database. We find that, with sufficient test and training data, suprasegmental information can significantly enhance the performance of traditional speaker ID systems.

show abstract

Section: Baseline Systems: Gmm and Lvcsrmentioning

confidence: 62%

“…Dragon Systems' GMM and LVCSR-based speaker ID systems are described in detail in [2]. We provide here an outline of their operation and performance on the Extended Data Task.…”

Section: Baseline Systems: Gmm and Lvcsrmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Using prosodic and lexical information for speaker identification

Weber¹,

Manganaro²,

Peskin

et al. 2002

IEEE International Conference on Acoustics Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…The segmentation step introduces an additional level of complexity in the speaker verification task. Speaker modeling with prior segment selection, based on large vocabulary continuous speech recognition (LVCSR), has already been studied [9,19,24,29]. The conclusions are that such systems provided state-of-the-art performance, given "sufficient" train and test material (2 min for training and 30 s for testing).…”

Section: Introductionmentioning

confidence: 97%

Segmental Approaches for Automatic Speaker Verification

Petrovska‐Delacrétaz

Černocký

Hennebert³

et al. 2000

Digital Signal Processing

View full text Add to dashboard Cite

“…As discussed in [2], Dragon traditionally uses three systems for speaker ID. The simplest is a basic Gaussian Mixture Model (GMM) system, which we treat as a baseline.…”

Section: Introductionmentioning

confidence: 99%

Speaker recognition in two-speaker data: recent results from Dragon Systems

Weber¹,

Peskin²,

Newman³

et al.

2000 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.00CH37100)

View full text Add to dashboard Cite

NIST has recently introduced two-speaker tasks in its Speaker Recognition evaluation scheme. We discuss Dragon Systems' recent explorations of these tasks using a GMM-based speaker ID system. We present baseline results submitted for the NIST two-speaker evaluation in April 1999. We also report on sonie "cheating" experiments and investigations of alternative segmentation and channel normalization techniques, which illuniinate some challenges specific t,o the two-speaker environment.

show abstract

Speaker Recognition on Single- and Multispeaker Data

Cited by 18 publications

References 13 publications

Using prosodic and lexical information for speaker identification

Using prosodic and lexical information for speaker identification

Segmental Approaches for Automatic Speaker Verification

Speaker recognition in two-speaker data: recent results from Dragon Systems

Contact Info

Product

Resources

About