Um Framework para Desenvolvimento de Sistemas TTS Personalizados no Português do Brasil

Ericson Sarmento Costa, Anderson de Oliveira Monte, Nelson Neto, Aldebaro Klautau

DOI: 10.14209/sbrt.2012.176

Evento: XXX Simpósio Brasileiro de Telecomunicações (SBrT2012)

Keywords: Hidden Markov Models HMM Speech Processing TTS.

Abstract

Recently in speech synthesis field many new results have been achieved using innovative techniques based on machine learning such as HMM-Based (Hidden Markov Models), which has generated good results in several languages by reducing the dificulty and time to build new voices for TTS (Text To Speech) systems. This paper aims to present a framework for building custom TTS systems in Brazilian Portuguese, with the main goal of being simple enough to be easily used by the community in general, besides being free to use.

Download