DocumentCode :
3529954
Title :
An analysis of articulatory-acoustic data based on articulatory strokes
Author :
Kato, Tsuneo ; Lee, Sungbok ; Narayanan, Shrikanth
Author_Institution :
Signal Anal. & Interpretation Lab., Univ. of Southern California, Los Angeles, SC
fYear :
2009
fDate :
19-24 April 2009
Firstpage :
4493
Lastpage :
4496
Abstract :
An articulatory gestural unit representation called ldquoarticulatory strokerdquo is introduced. It aims to capture the constriction formation and release of a speech articulator such as the tongue tip. In that regard, the articulatory stroke is an attempt at a practical realization of the abstract articulatory gestures central to Articulatory Phonology. In this study we focus on the articulatory strokes associated with the critical articulator that is essential to realize a target phone. The critical articulatory stroke is parameterized in order to investigate the predictability of the parameters from phonetic contexts and to check the statistical dependency of acoustic changes associated with the critical articulatory strokes. Canonical correlation analysis between the articulatory strokes and DeltaMFCCs showed that the critical articulatory strokes are more responsible to the acoustic changes inside target phonemes than non-critical articulators. This implies that modeling acoustic changes due to critical articulations could provide an edge in improving ASR performance.
Keywords :
correlation methods; speech processing; speech recognition; articulatory acoustic data; articulatory gestural unit representation; articulatory gestures; articulatory phonology; articulatory stroke; canonical correlation analysis; constriction formation; critical articulations; speech articulator; tongue tip; Automatic speech recognition; Data analysis; Data engineering; Hidden Markov models; Laboratories; Research and development; Signal analysis; Speech analysis; Tongue; Viterbi algorithm; articulatory stroke; automatic speech recognition; canonical correlation analysis; critical articulation;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Acoustics, Speech and Signal Processing, 2009. ICASSP 2009. IEEE International Conference on
Conference_Location :
Taipei
ISSN :
1520-6149
Print_ISBN :
978-1-4244-2353-8
Electronic_ISBN :
1520-6149
Type :
conf
DOI :
10.1109/ICASSP.2009.4960628
Filename :
4960628
Link To Document :
بازگشت