This paper reports on recent developments of a research on nonlinear methods in speech analysis and on flexible modelling of speech production. The final aim of our research is to bring together speech analysis and synthesis, in order to better understand the underlying processes of speech and to address specific applications. Several techniques used in the analysis of dynamic nonlinear systems, are applied in order to investigate some of the short-term nonlinear characteristics of speech signal production. The research starts from the physiological evidence that the phonatory system is a time-varying nonlinear system, moreover that nonlinear processes are present in speech production, such as turbulent flow. Before the main analysis, the speech signal is decomposed into two parts: a low-dimensional, almost linear part and a high-dimensional nonlinear part, respectively. For the latter, the largest Lyapunov exponent and fractal dimensions (capacity, correlation and information) are computed. The synthesis tool is based on a fuzzy model of speech signal production, that implements knowledge about the speech production apparatus.
speech production, nonlinear and nonstationary dynamic processes, Lyapunov exponents, fuzzy modelling, source-filter paradigm.
All articles published by Studies in Informatics and Control (SIC) are published in accordance with the terms and conditions of the Creative Commons Attribution License (CC BY-NC-SA 4.0).