Free tools and resources for HMM-based Brazilian Portuguese speech synthesis

Abstract

Text-to-speech (TTS) is currently a mature technology used in many areas such as education and accessibility. Some modules of a TTS system depend on the language and, while there are many public materials for some languages (e.g., English and Japanese), the resources for Brazilian Portuguese (BP) are still limited. This work describes the development of a complete hidden Markov model (HMM) based TTS system for BP which can be applied to the desktop environment. It also releases a set of natural language processing tools for BP, which expands the already publicly available resources, supporting the development of new researches for academic or industrial purposes. Subjective and objective performance tests are presented, comparing the proposed TTS system with other softwares currently available for BP.