Text-to-Speech solution based on MBROLA Engine for developing input data forms
Supp/2007 , p 725..733
Author(s):
Traian SURCEL
Cristian PETRE
Keywords:
Text-to-speech, Mbrola - engine, computer-human interface, information system
Abstract:
The TTS technologies transform a sequence of words in sounds by the synthesis of the human voice. We considered that the TTS technologies are useful to improve the quality the computer - human interface in management information systems. We developed a software solution based on MBROLA Synthesizer and VB programs for building of the input data forms. MBROLA - Multilingual Speech Synthesizer is free software with a Romanian language phonetic units data base. MBROLA uses units’ selection concatenation as the method of generate and playing sounds. First it must find and extract the diphone units from the MBROLA data base using the phonetic description of the text. After this, using also the prosody information linked to the diphone, such as duration of the phonemes and linear description of the intonation of sounds, the MBROLA Engine plays the words. The VB–MBROLA interface performs in the first stage the front-end processing of the text, the text normalization and generating of the text-to-diphone representation. In the second stage, the back-end processing, VB program uses the MBRALA functions and a phoneme-player to generate vocal messages. The user answers to vocal messages by entering input data required by the form designed with VB.
Download:
Back
|
|