Title: | Robust statistic estimates for adaptation in the task of speech recognition |
Other Titles: | Robustní odhad statistik pro adaptaci v úloze rozpoznávání řeči |
Authors: | Zajíc, Zbyněk Machlica, Lukáš Müller, Luděk |
Citation: | ZAJÍC, Zbyněk; MACHLICA, Lukáš; MÜLLER, Ludě›k. Robust statistic estimates for adaptation in the task of speech recognition. In: Lecture notes in computer science. Berlin: Springer, 6231/2010, p. 464-471. ISSN 0302-9743 |
Issue Date: | 2010 |
Publisher: | Springer |
Document type: | článek article |
URI: | http://www.kky.zcu.cz/cs/publications/ZbynekZajic_2010_RobustStatistic http://hdl.handle.net/11025/16953 |
ISSN: | 0302-9743 |
Keywords: | fMLLR;adaptace;rozpoznávání řeči;robustnost |
Keywords in different language: | fMLLR;adaptation;speech recognition;robustness |
Abstract: | Tento článek se zabývá robustním odvozením statistik pro úlohu adaptace akustického modelu. Statistiky jsou akumulovány před vlastním procesem adaptace z dostupných adaptačních dat. Obecně se předpokládá malé množství těchto dat, která jsou často znehodnocena šumem, různým kanálem atd., proto jsme navrhli techniku, která má za úkol robustní odhat adaptace z takovýchto dat. Jednou z často užívaných metod je inicializace adaptačních tranformací. Druhá, námi navržená metoda spočívá v akumulování pouze vhodně rozpoznaných dat (statistik). Zameřili jsme se na metodu fMLLR, na které jsme provedli experimenty. |
Abstract in different language: | This paper deals with robust estimations of data statistics used for the adaptation. The statistics are accumulated before the adaptation process from available adaptation data. In general, only small amount of adaptation data is assumed. These data are often corrupted by noise, channel, they do not contain only clean speech. Also, when training Hidden Markov Models (HMM) several assumptions are made that could not have been fulfilled in the praxis, etc. Therefore, we described several techniques that aim to make the adaptation as robust as possible in order to increase the accuracy of the adapted system. One of the methods consists in initialization of the adaptation statistics in order to prevent ill-conditioned transformation matrices. Another problem arises when an acoustic feature is assigned to an improper HMM state even if the reference transcription is available. Such situations can occur because of the forced alignment process used to align frames to states. Thus, it is quite handy to accumulate data statistic utilizing only reliable frames (in the sense of data likelihood). We are focusing on Maximum Likelihood Linear Transformations and the experiments were performed utilizing the feature Maximum Likelihood Linear Regression (fMLLR). Experiments are aimed to describe the behavior of the system extended by proposed methods. |
Rights: | © Zbyněk Zajíc - LukášMachlica - Luděk Müller |
Appears in Collections: | Články / Articles (KKY) Články / Articles (NTIS) |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
ZbynekZajic_2010_RobustStatistic.pdf | Plný text | 166,99 kB | Adobe PDF | View/Open |
Please use this identifier to cite or link to this item:
http://hdl.handle.net/11025/16953
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.