Imitação da voz humana através do processo de análise-por-síntese utilizando algoritmo genético e sintetizador de voz por formantes

Abstract

Voice imitation through the utterance copy mechanism is estimating the value of the input parameters of a speech synthesizer to generate a similar signal with the original voice. This process is distinct from the more traditional text-to-speech, but yet used in many areas, especially, Linguistics and Health System. Imitate the human voice through this mechanism is a difficult inverse problem because the mapping is non-linear and from many to one. For instance, there are different combinations of the synthesizer input parameters values that produce the same synthetic voice signal. Therefore, perform voice imitation manually requires a considerable amount of time. In addition to automatic methods are our interest of study as well, as proposed here. This work presents our system based on Genetic Algorithm (GA) to automatically estimate the value of the input parameters of a speech formant synthesizer using the analysis-by-synthesis process. Results are presented for synthetic (computer-generated) and natural (human-generated) speech in American English, for male and female speakers. These results are compared with the ones obtained with Winsnoori, the only currently available software that performs the same task. The experiments showed that the proposed newGASpeech framework is an effective alternative to the laborious manual process of estimating the input parameters values of a formant synthesizer. Besides it has overcome the quality of the generated voices by the baseline if compared to five objective metrics and a subjective evaluation applied to twenty seven no-expert listeners in the speech area neither the adopted language.