| Title: | The speaker adaptation of an acoustic model |
| Other Titles: | Adaptace akustického modelu |
| Authors: | Machlica, Lukáš Zajíc, Zbyněk |
| Citation: | MACHLICA, Lukáš; ZAJÍC, Zbyněk. The speaker adaptation of an acoustic model. In: The 1st Young Researchers Conference on Applied Sciences: conference proceedings book: 13 November 2007 University of West Bohemia. PlzeĹ: ZápadoÄŤeská univerzita v Plzni, 2007, s. 212-217. ISBN 978-80-7043-574-8. |
| Issue Date: | 2007 |
| Publisher: | Západočeská univerzita v Plzni |
| Document type: | článek article |
| URI: | http://www.kky.zcu.cz/cs/publications/MachlicaL_2007_TheSpeaker http://hdl.handle.net/11025/17037 |
| ISBN: | 978-80-7043-574-8 |
| Keywords: | adaptace;MAP;MLLR;CMLLR;combination |
| Keywords in different language: | adaptation;MAP;MLLR;CMLLR;combination |
| Abstract: | Tento článek popisuje několik adaptačních metod. Každá z metod přináší různé výhody.Zabývali jsme se i jejich kombinacemi. Byly použity metody MAP, MLLR a CMLLR. Všechny modely jsou založeny na HMM přístupu. |
| Abstract in different language: | This paper deals with several adaptation techniques, which are of the importance in cases when the identity of a speaker is known and we want to recognize his speech. Each of the methods yields various benefits, therefore we examined their combination. This approach brought further error rate decreasing. We are using three different methods, namely Maximum Apriori Probability adaptation, Maximum Likelihood Linear Regression and Constrained Maximum Likelihood Linear Regression. All acoustic models are based on the Hiden Markov Model. |
| Rights: | © Západočeská univerzita v Plzni |
| Appears in Collections: | Články / Articles (KKY) Články / Articles (NTIS) |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| MachlicaL_2007_TheSpeaker.pdf | Plný text | 110,94 kB | Adobe PDF | View/Open |
Please use this identifier to cite or link to this item:
http://hdl.handle.net/11025/17037Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.