Home // ICDT 2020, The Fifteenth International Conference on Digital Telecommunications // View article

A Model for Infant Acquisition of Spoken Words Using Genetic Algorithm and Fujisaki Model

Authors:
Tomio Takara
Ryoichi Eto

Keywords: model for acquisition of spoken word; coarticulation; Fujisaki model; genetic algorithm; vowel space parameter

Abstract:
We propose a new model of speech imitation and acquisition process of infants. We regard the vowel space parameters as the articulatory gesture, such as the tongue hump position and the degree of constriction of vowels. We represent the coarticulation effect using Fujisaki’s generative model of speech. We model a trial and error process of the infant’s speech imitation using the Genetic Algorithm (GA). In our model, we regard “command” in the Fujisaki model as the articulatory gesture and detect it from the spectral sequence using the GA. In other words, the original phonemic target is inversely estimated as the Fujisaki’s command from the phonemically ambiguous speech spectrum caused by the coarticulation. Our model simulates the hypothesis that human infants acquire the normalization (inverse estimation) skill of the coarticulation through the process of imitating spoken words. We evaluated the model in listening tests using synthesized speech. We also show that the model can represent the phenomenon of “predicted sound”, which is unconsciously heard as the effect of the normalization of the coarticulation, by comparing this predicted sound with the inversely estimated sound.

Pages: 1 to 6

Copyright: Copyright (c) IARIA, 2020

Publication date: February 23, 2020

Published in: conference

ISSN: 2308-3964

ISBN: 978-1-61208-768-9

Location: Lisbon, Portugal

Dates: from February 23, 2020 to February 27, 2020