Development and evaluation of PATRICIA

a real-time singing synthesizer for the Brazilian Portuguese language

Abstract

Singing voice synthesis involves the artificial generation of a song given its lyrics and musical notes through computational methods. This paper describes the development and evaluation process of PATRICIA, a proof-of-concept prototype of a system that performs real-time singing voice synthesis (SVS) for the Brazilian Portuguese language. It is the first singing synthesizer specifically designed for this idiom. A literature review was conducted to study the latest developments in real-time singing synthesis and to give directions for PATRICIA design and implementation. The implemented architecture is described, and, for the validation process, a series of video recordings of the system working was made as a musical demonstration. This demonstration was subjected to an evaluation by musical educators. A performance analysis with two CPU usage indicators in two different devices was also conducted. The results of the validation are discussed, and future enhancements are pointed out in order to overcome the system’s current limitations.