Desarrollo de un prototipo de convertidor de texto a voz por medio de síntesis concatenativa para realizar análisis comparativo entre los métodos MBR-PSOLA y FD-PSOLA

In this work, a prototype of text to speech converter is developed through the MATLAB programming language, which uses synthesis for concatenation. For this, four important stages are performed: first, the definition of the acoustic, which depends on the type of concatenated synthesis that is to be performed, second, the creation of the voice corpus for the Spanish language, third, the creation of the natural language processing module, which performs the analysis of all input text and finally the creation of the signal processing module, where the whole process of synthesis is carried out to obtain the final audio. In the signal processing step, a module of synthesis is added, which implement the FD-PSOLA and MBR-PSOLA methods, to improve the co-articulation step. Having already done the converter from text to speech we proceed to perform comparative tests, both subjective and objective, which will allow to evaluate the vocal quality of this synthesizer depending on the method used.

• Autores: Mususué Castro, Vanessa; Rodríguez Vásquez, Julián Andrés
• Fecha de publicación: 2017
• Institución: Universidad de San Buenaventura
• Repositorio: Repositorio USB
• Idioma: spa
• OAI Identifier: oai:bibliotecadigital.usb.edu.co:10819/5732
• Acceso en línea: http://hdl.handle.net/10819/5732
• Palabra clave: Convertidor texto a voz; Síntesis del habla; Coarticulación; Difonemas; Naturalidad; PSOLA, FD-PSOLA; Text-to-speech converter; Speech synthesis; Co-articulation; Diphoneme; Naturalness; Convertidores; Lenguaje; Acústica; Síntesis - sonido
• License: Atribución-NoComercial-SinDerivadas 2.5 Colombia