Abstract
Many biomedical systems exhibit some form of nonlinear behavior, such as saturation in response to certain inputs. Linear description may not be sufficient to describe the complex nonlinear phenomena, including bifurcation and chaos. The determination of nonlinearity from time series has become an important topic for the analysis and modeling of biomedical systems. In this study, we will apply nonlinear dynamic modeling based on Volterra kernel function to describe the nonlinearity of voice signals. Three nonlinear parameters including nonlinearity-to-linearity amplitude ratio (NLAR), nonlinearity-to-linearity parameter ratio (NLPR), and nonlinear order are employed. Sufficiently lower NLAR and NLPR, and a nonlinear order approaching 1 can be found in periodic voice signals, demonstrating the linear mechanism of periodic voice production. However, aperiodic voices show sufficiently high NLAR and NLAR, and a nonlinear order above 2, indicating the dominant role of nonlinearity in disordered voice production. Furthermore, the effects of the signal length and noises on these three parameters are investigated. These three nonlinear parameters show the robustness of short signal length and noise perturbations, demonstrating their potential applications in measuring the nonlinearity of disordered voice production systems. [Work supported by NIH.]
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call