Attributes | Values |
---|
rdf:type
| |
Description
| - One of the most utilized adaptation techniques is the feature Maximum Likelihood Linear Regression (fMLLR). In comparison with other adaptation methods the number of free parameters to be estimated significantly decreases. Thus, the method is well suited for situations with small amount of adaptation data. However, fMLLR still fails in situations with extremely small data sets. Such situations can be solved through proper initialization of fMLLR estimation adding some a-priori information. In this paper a novel approach is proposed solving the problem of fMLLR initialization involving statistics from speakers acoustically close to the speaker to be adapted. Proposed initialization suitably substitutes missing adaptation data with similar data from a training database, fMLLR estimation becomes well-conditioned, and the accuracy of the recognition system increases even in situations with extremely small data sets.
- One of the most utilized adaptation techniques is the feature Maximum Likelihood Linear Regression (fMLLR). In comparison with other adaptation methods the number of free parameters to be estimated significantly decreases. Thus, the method is well suited for situations with small amount of adaptation data. However, fMLLR still fails in situations with extremely small data sets. Such situations can be solved through proper initialization of fMLLR estimation adding some a-priori information. In this paper a novel approach is proposed solving the problem of fMLLR initialization involving statistics from speakers acoustically close to the speaker to be adapted. Proposed initialization suitably substitutes missing adaptation data with similar data from a training database, fMLLR estimation becomes well-conditioned, and the accuracy of the recognition system increases even in situations with extremely small data sets. (en)
|
Title
| - Initialization of fMLLR with Sufficient Statistics from Similar Speakers
- Initialization of fMLLR with Sufficient Statistics from Similar Speakers (en)
|
skos:prefLabel
| - Initialization of fMLLR with Sufficient Statistics from Similar Speakers
- Initialization of fMLLR with Sufficient Statistics from Similar Speakers (en)
|
skos:notation
| - RIV/49777513:23520/11:43898196!RIV12-GA0-23520___
|
http://linked.open...avai/riv/aktivita
| |
http://linked.open...avai/riv/aktivity
| - P(GA102/08/0707), P(LC536), S
|
http://linked.open...iv/cisloPeriodika
| |
http://linked.open...vai/riv/dodaniDat
| |
http://linked.open...aciTvurceVysledku
| |
http://linked.open.../riv/druhVysledku
| |
http://linked.open...iv/duvernostUdaju
| |
http://linked.open...titaPredkladatele
| |
http://linked.open...dnocenehoVysledku
| |
http://linked.open...ai/riv/idVysledku
| - RIV/49777513:23520/11:43898196
|
http://linked.open...riv/jazykVysledku
| |
http://linked.open.../riv/klicovaSlova
| - fMLLR, adaptation, sufficient statistics, speech recognition, robustness,initialization. (en)
|
http://linked.open.../riv/klicoveSlovo
| |
http://linked.open...odStatuVydavatele
| - DE - Spolková republika Německo
|
http://linked.open...ontrolniKodProRIV
| |
http://linked.open...i/riv/nazevZdroje
| - Lecture Notes in Computer Science
|
http://linked.open...in/vavai/riv/obor
| |
http://linked.open...ichTvurcuVysledku
| |
http://linked.open...cetTvurcuVysledku
| |
http://linked.open...vavai/riv/projekt
| |
http://linked.open...UplatneniVysledku
| |
http://linked.open...v/svazekPeriodika
| |
http://linked.open...iv/tvurceVysledku
| - Machlica, Lukáš
- Zajíc, Zbyněk
- Müller, Luděk
|
issn
| |
number of pages
| |
http://bibframe.org/vocab/doi
| - 10.1007/978-3-642-23538-2_24
|
http://localhost/t...ganizacniJednotka
| |