This HTML5 document contains 48 embedded RDF statements represented using HTML+Microdata notation.

The embedded RDF content will be recognized by any processor of HTML5 Microdata.

Namespace Prefixes

PrefixIRI
dctermshttp://purl.org/dc/terms/
n20http://localhost/temp/predkladatel/
n14http://linked.opendata.cz/resource/domain/vavai/projekt/
n7http://linked.opendata.cz/resource/domain/vavai/riv/tvurce/
n11http://linked.opendata.cz/resource/domain/vavai/subjekt/
n10http://linked.opendata.cz/ontology/domain/vavai/
shttp://schema.org/
skoshttp://www.w3.org/2004/02/skos/core#
n4http://linked.opendata.cz/ontology/domain/vavai/riv/
n13http://linked.opendata.cz/resource/domain/vavai/vysledek/RIV%2F49777513%3A23520%2F11%3A43898196%21RIV12-GA0-23520___/
n19http://bibframe.org/vocab/
n2http://linked.opendata.cz/resource/domain/vavai/vysledek/
rdfhttp://www.w3.org/1999/02/22-rdf-syntax-ns#
n12http://linked.opendata.cz/ontology/domain/vavai/riv/klicoveSlovo/
n18http://linked.opendata.cz/ontology/domain/vavai/riv/duvernostUdaju/
xsdhhttp://www.w3.org/2001/XMLSchema#
n15http://linked.opendata.cz/ontology/domain/vavai/riv/aktivita/
n6http://linked.opendata.cz/ontology/domain/vavai/riv/jazykVysledku/
n17http://linked.opendata.cz/ontology/domain/vavai/riv/obor/
n16http://linked.opendata.cz/ontology/domain/vavai/riv/druhVysledku/
n5http://reference.data.gov.uk/id/gregorian-year/

Statements

Subject Item
n2:RIV%2F49777513%3A23520%2F11%3A43898196%21RIV12-GA0-23520___
rdf:type
skos:Concept n10:Vysledek
dcterms: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.
dcterms:title
Initialization of fMLLR with Sufficient Statistics from Similar Speakers Initialization of fMLLR with Sufficient Statistics from Similar Speakers
skos:prefLabel
Initialization of fMLLR with Sufficient Statistics from Similar Speakers Initialization of fMLLR with Sufficient Statistics from Similar Speakers
skos:notation
RIV/49777513:23520/11:43898196!RIV12-GA0-23520___
n10:predkladatel
n11:orjk%3A23520
n4:aktivita
n15:S n15:P
n4:aktivity
P(GA102/08/0707), P(LC536), S
n4:cisloPeriodika
6836
n4:dodaniDat
n5:2012
n4:domaciTvurceVysledku
n7:8612889 n7:3020614 n7:6895972
n4:druhVysledku
n16:J
n4:duvernostUdaju
n18:S
n4:entitaPredkladatele
n13:predkladatel
n4:idSjednocenehoVysledku
204873
n4:idVysledku
RIV/49777513:23520/11:43898196
n4:jazykVysledku
n6:eng
n4:klicovaSlova
fMLLR, adaptation, sufficient statistics, speech recognition, robustness,initialization.
n4:klicoveSlovo
n12:initialization. n12:adaptation n12:fMLLR n12:sufficient%20statistics n12:speech%20recognition n12:robustness
n4:kodStatuVydavatele
DE - Spolková republika Německo
n4:kontrolniKodProRIV
[1CEB7BC95AF0]
n4:nazevZdroje
Lecture Notes in Computer Science
n4:obor
n17:JD
n4:pocetDomacichTvurcuVysledku
3
n4:pocetTvurcuVysledku
3
n4:projekt
n14:LC536 n14:GA102%2F08%2F0707
n4:rokUplatneniVysledku
n5:2011
n4:svazekPeriodika
Neuveden
n4:tvurceVysledku
Zajíc, Zbyněk Müller, Luděk Machlica, Lukáš
s:issn
0302-9743
s:numberOfPages
8
n19:doi
10.1007/978-3-642-23538-2_24
n20:organizacniJednotka
23520