DocumentCode :
2580588
Title :
Vocal tract length normalization strategy based on maximum likelihood criterion
Author :
Jakovljevic, Niksa M. ; Secujski, M.S. ; Delic, V.D.
fYear :
2009
fDate :
18-23 May 2009
Firstpage :
399
Lastpage :
402
Abstract :
In this paper performances of automatic speech recognition systems which use vocal tract length normalization (VTN) are presented. Beside standard procedure for VTN coefficient estimation several variants based on robust statistic methods are introduced. All systems which use VTN performed better than referent systems, while the best performance was achieved by the system in which the VTN coefficient for a particular speaker is chosen as the one with maximum sample mean of likelihoods per phoneme. Phoneme likelihoods are calculated as sample medians of feature vectors corresponding to particular phonemes. The relative improvement of performance for this system is about 20%.
Keywords :
maximum likelihood estimation; speech recognition; automatic speech recognition systems; coefficient estimation; feature vectors; maximum likelihood criterion; particular speaker; phoneme likelihoods; robust statistic methods; vocal tract length normalization strategy; Automatic speech recognition; Hidden Markov models; Loudspeakers; Materials testing; Maximum likelihood estimation; Mel frequency cepstral coefficient; Parameter estimation; Piecewise linear techniques; Robustness; Statistics; speech recognition; vocal tract length normalization;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
EUROCON 2009, EUROCON '09. IEEE
Conference_Location :
St.-Petersburg
Print_ISBN :
978-1-4244-3860-0
Electronic_ISBN :
978-1-4244-3861-7
Type :
conf
DOI :
10.1109/EURCON.2009.5167662
Filename :
5167662
Link To Document :
بازگشت