A system for automatic labeling and segmentation of speech signals starting from their corresponding text will be described. The system uses continuous Hidden Markov Models (HMM) to represent a predefined set of acoustic-phonetic units and pronunciation networks to allow different phonetic realizations of a given sentence. The system has been applied to an American (TIMIT) and an Italian (APASCI) speech database
Automatic Speech Labeling using Word Pronunciation Networks and Hidden Markov Models
Angelini, Bianca;Brugnara, Fabio;Giuliani, Diego;Falavigna, Giuseppe Daniele;Gretter, Roberto;Omologo, Maurizio
1995-01-01
Abstract
A system for automatic labeling and segmentation of speech signals starting from their corresponding text will be described. The system uses continuous Hidden Markov Models (HMM) to represent a predefined set of acoustic-phonetic units and pronunciation networks to allow different phonetic realizations of a given sentence. The system has been applied to an American (TIMIT) and an Italian (APASCI) speech databaseFile in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
