Sistema de conversão texto-fala para a lingua portuguesa utilizando a abordagem de sintese por regras

AUTOR(ES)
DATA DE PUBLICAÇÃO

1998

RESUMO

This work contains a description of the text-to-speech conversion system for the Portuguese of Brazil developed at the Signal Processing Laboratory of the Electrical and Computer Engineering School of Unicamp. The system receives as input a generic text in Portuguese and produces as output the corresponding speech signal. The text-to-speech conversion process is divided into three basic steps, each one including several modules: Text processing: preprocessing, grammatical classification, syllabic division and orthographic-phonetic transcription. Prosodic processing: determination of prosodic boundaries, generation of intonation patterns and generation of segmental durations. Signal processing: synthesis of the speech signal using the Klatt formant synthesizer. The prosodic processing modules use duration and intonation data extracted from natural utterances, adjusting them to the particularities of the input text on the basis of information provided by the grammatical classifier. The synthesis-by-rule approach is used for generating the synthesizer control parameters. A specific language and a compiler have been created for the description of the synthesis rules. A1thoughformal evaluations of the system quality have not been made, informal tests have indicated a good general performance in terms of intelligibility and naturalness

ASSUNTO(S)

sintese da voz sistemas de processamento da fala

ACESSO AO ARTIGO

Documentos Relacionados