Name
Abstract | ||
|---|---|---|
In this paper we study the problem of identifying in-set and out- of-set speakers. The goal is to identify whether an unknown input speaker belongs to either a group of in-set speaker or an unseen out-of-set group. A state-of-the-art GMM classifier with Universal Background Model (UBM), and standard likelihood ratio test are used as our baseline system. We propose an alternative hypothesis testing method that employs neighborhood information with re- spect to each in-set speaker model in the model space based on the Kullback-Leibler divergence. The Bayes Factor is used in the veri- fication stage (accept/reject hypothesis). We evaluate the proposed procedure on a clean CORPUS1 set, and a noisy CORPUS2 set which contains session-to-session variability. Experiments show an improvement in Equal Error Rate for the system even when in- set speaker models are acoustically close in the model space, and as the in-set speaker size increases. |
Year | DOI | Venue |
|---|---|---|
2004 | 10.1109/ICASSP.2004.1326005 | ICASSP '04). IEEE International Conference |
Keywords | Field | DocType |
Bayes methods,Gaussian processes,error statistics,speaker recognition,Bayes factor,GMM classifier,Kullback-Leibier divergence,equal error rate,hypothesis testing method,in-set speaker identification,neighborhood information,out-of-set speaker identification,standard likelihood ratio test,universal background model | Alternative hypothesis,Pattern recognition,Likelihood-ratio test,Computer science,Word error rate,Bayes factor,Speech recognition,Speaker recognition,Artificial intelligence,Speaker diarisation,Classifier (linguistics),Bayes error rate | Conference |
Volume | ISSN | ISBN |
1 | 1520-6149 | 0-7803-8484-9 |
Citations | PageRank | References |
4 | 0.52 | 6 |
Authors | ||
2 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Pongtep Angkititrakul | 1 | 179 | 15.47 |
| John H. L. Hansen | 2 | 3215 | 365.75 |