Nonlinear Dynamical Detection of Normal Vowels Using PPS Algorithm and WDP of Neural Networks

Normal vowels are confirmed to have irregular property which is possibly due to chaos. In this paper, we employ two approaches to detect the existence of underlying dynamics in Chinese normal vowels recorded from both male and female adults. The PPS algorithm that can test pseudoperiodic time series against the hypothesis of periodic orbits with uncorrelated noise presents that these data sets follow chaos. Meanwhile, another novel technique, weight distribution projection (WDP) of neural networks, is proposed to identify dynamical property of these given time series. Intuitively, their WDP graphs exhibit the chaotic property. So both methods give the same conclusion that the measured Chinese vowels are consistent with chaos.