Name
Abstract | ||
|---|---|---|
Model migration in speaker recognition is a task of convert- ing parametrically-obsolete models to new structures and configurations without the requirement to store the orig- inal speech waveforms or feature vector sequences along with the models. The need for model migration arises in large-scale deployments of speaker recognition technology in which the potential for legacy problems increases as the evolving technology may require configuration changes thus invalidating already existing user voice accounts. A migra- tion may represent the only alternative to otherwise costly user re-enrollment or waveform storage and, as a new re- search problem, presents the challenge of developing algo- rithms to minimize the loss in accuracy in the migrated accounts. This paper reports on further enhancements of a statistical migration technique based on Gaussian Mixture Models, introduced previously. The present approach is based on a stochastic synthesis of feature sequences from ob- solete models that are subsequently used to create the new models. Here, in addition to Gaussian means and priors, as utilized in the previous contribution, also the covariances are included resulting in significant performance gains in the migrated models, compared to the mean-only method. Overall, measured on the NIST 2003 cellular task, the de- scribed algorithm achieves a model migration incurring a loss in performance of 8-20% relative to a full re-enrollment from waveforms, dependent on the type of mismatch be- tween the obsolete and the new configuration. The inclu- sion of the covariance information is shown to reduce the loss of performance by a factor of 3-4 as compared to the baseline mean-only migration technique. |
Year | DOI | Venue |
|---|---|---|
2005 | 10.1109/ICASSP.2005.1415193 | ICASSP '05). IEEE International Conference |
Keywords | DocType | Volume |
Gaussian distribution,covariance analysis,speaker recognition,stochastic systems,Gaussian covariances,Gaussian means,Gaussian mixture models,Gaussian priors,MAP estimation,MAP-based algorithm,feature sequence stochastic synthesis,speaker model migration,speaker recognition,statistical migration technique | Conference | 1 |
ISSN | ISBN | Citations |
1520-6149 | 0-7803-8874-7 | 0 |
PageRank | References | Authors |
0.34 | 5 | 2 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Jiri Navratil | 1 | 314 | 31.36 |
| Ganesh N. Ramaswamy | 2 | 213 | 25.72 |