Dissertation

 
Schwarz, P.: Phoneme recognition based on long temporal context, Brno, CZ, FIT VUT, 2009, p. 95
Publication language:english
Original title:Phoneme recognition based on long temporal context
Title (cs):Rozpoznávání fonémů založené na dlouhém časovém kontextu
Pages:95
Place:Brno, CZ
Year:2009
Publisher:Faculty of Information Technology BUT
URL:http://www.fit.vutbr.cz/research/groups/speech/publi/2009/schwarz_thesis.pdf [PDF]
Keywords
phoneme recognition, TIMIT, neural networks, temporal patterns, long temporal context, split temporal context, language identification
Annotation
The work is on phoneme recognition based on long temporal context
Abstract
Techniques for automatic phoneme recognition from spoken speech are investigated. The goal is to extract as much information about phoneme from as long temporal context as possible. The Hidden Markov Model / Artificial Neural Network (HMM/ANN) hybrid system is used. At first, the Temporal Pattern (TRAP) system is implemented and compared to other systems based on conventional feature extraction techniques. The TRAP system is analyzed and simplified. Then a new Split Temporal Context (STC) system is proposed. The system reach better results while the complexity was reduced. Then the system was improved using commonly used techniques like three-state phoneme modelling and phonotactic language model. Such system reaches 21.48% phoneme error rate on the TIMIT database. The STC system was also studied on another databases, in noise and in cross-channel conditions. Finally few applications where the phoneme recognizer was applied are demonstrated.
BibTeX:
@PHDTHESIS{
   author = {Petr Schwarz},
   title = {Phoneme recognition based on long temporal context},
   pages = {95},
   year = {2009},
   location = {Brno, CZ},
   publisher = {Faculty of Information Technology BUT},
   language = {english},
   url = {http://www.fit.vutbr.cz/research/view_pub.php?id=9132}
}