GASpeech: A framework for automatically estimating input parameters of Klatt's speech synthesizer

This work describes GASpeech: a framework centered on genetic algorithms for automatically estimating the input parameters of Klatt's speech synthesizer. GASpeech aims to speed up the process of speech imitation (or utterance copy), where one has to find the model parameters that lead to a synthesized speech sounding close enough to the natural target speech (i.e. low spectral distortion). The architecture of GASpeech is described, emphasizing the usage of adaptive control of probabilities and stopping criteria speeding up the convergence process. Though this paper does not present a closed solution for all Klatt's parameters, an enormous improvement in accuracy and methodology was reached when compared to previous works.