Title | ||
---|---|---|
Semi-supervised speech activity detection with an application to automatic speaker verification. |
Abstract | ||
---|---|---|
•We propose a new speech activity detector (SAD) based on semi-supervised learning of Gaussian mixture model (GMM).•The proposed SAD requires lower amount of data labeled data for initialization as compared to GMM-based approach.•We have shown improved detection of speech and non-speech frames on NIST OpenSAD dataset.•The proposed SAD gives promising results compared to other SADs in robust speaker verification task.
|
Year | DOI | Venue |
---|---|---|
2018 | 10.1016/j.csl.2017.07.005 | Computer Speech & Language |
Keywords | Field | DocType |
Speech activity detection,Semi-supervised learning,Gaussian mixture model,Speaker recognition,NIST OpenSAD,NIST SRE | Semi-supervised learning,Computer science,Speaker recognition,Speaker diarisation,Artificial intelligence,Pattern recognition,Voice activity detection,Word error rate,Speech recognition,NIST,Initialization,Machine learning,Mixture model | Journal |
Volume | Issue | ISSN |
47 | C | 0885-2308 |
Citations | PageRank | References |
4 | 0.41 | 26 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Alexey Sholokhov | 1 | 4 | 0.41 |
Md. Sahidullah | 2 | 326 | 24.99 |
Tomi Kinnunen | 3 | 1323 | 86.67 |