Chaotic dynamics and simulation of Japanese vowel sounds

We examine deterministic dynamics in Japanese speech vocalisation patterns. We find that a simple data driven non-parametric model may be employed to accurately and compactly describe the complex dynamic structure of the data. Moreover, we demonstrate that the data is not consistent with a monotonic nonlinear transformation of linearly filtered noise (i.e. determinism in the data cannot be adequately described by the Fourier power spectrum). We then show that for four of the five vowel sounds, the dynamics are not consistent with a noise driven nonlinear periodic orbit, and therefore the underlying dynamical system is likely to be chaotic. In the fifth case (the vowel /a/) we are unable to reject the hypothesis of a noise driven periodic orbit.