Nonlinear Analysis Of Pathological Phonation Model And Parameter Optimization

Voice quality directly affects the communication skills. In recent years, the prevalence of vocal disorders increases every year due to the environment, occupation and other factors, which has greatly affected the quality of life. Acoustic detection has been widely used in pathological voice recognition, but the study of characteristic parameters is still deficient, and the selection of parameters is limited to acoustic parameters. Vocal cord lesion is an important reason of abnormal voice. In recent years, according to the structure and mechanism of vocal sound organization, researchers established various physical models to simulate the vocal cords vibration, study on the parameters related to the phonation model has provide a new way for vocal diseases classification.Against the limitations of the existing acoustic parameters, continuous nonlinear analysis is proposed on the phonation physical model. The model parameters are optimized to match the actual voice source. Both the difference of the continuous nonlinear parameters and physical parameters extracted from the model are analyzed. Firstly, we analyze the vibration of normal and pathological vocal system. Considering the influence of sub-glottal pressure on the vocal system, parameters related to fundamental frequency,amplitude and nonlinear dynamics parameters are extracted under the condition of different sub-glottal pressure. Continuous nonlinear parameters under different sub-glottal pressure and their difference are analyzed. Then the model is optimized to match the actual voice source. Difference analysis is done between the model parameters of different diseases.Specific studies of the thesis are as follows:(1) According to the physical structure of the vocal cords, we establish the model of normal or pathological vocal cord which is coupled to sub-glottal airflow to generate voicesource. The fundamental frequency and related perturbation parameters are extracted.Poincare section and bifurcation diagram are proposed to analyze the vibration. Lastly, the pathological parameters and sub-glottal pressure are changed to analyze the changes in the frequency parameters and chaos parameter such as Lyapunov exponent. The simulation results show that vocal cord paralysis reduces the fundamental frequency, in which the chaos occurs only within a certain pressure range; while vocal cord with polyps do not reduce the fundamental frequency, in which the chaos is distributed throughout the entire range.(2) In order to adjust the model parameters to match the actual vocal parameters, and to describe the physiological parameters changing caused by vocal laryngeal lesions,particle swarm optimization combined with Quasi-Newton algorithm is proposed to optimize the model. As the actual voice source is accurately simulated with the optimized model, we get the parameters that could represent the actual vocal cord structure. Then difference analysis is done between the optimized model parameters.This thesis introduces physical models of the Pathological laryngeal voice source,proposes the maximum Lyapunov exponent under continuous sub-glottal pressure, and optimizes the model parameters further. The results show that there are significant differences between vocal cord paralysis and vocal polyps according to the largest Lyapunov exponent distribution. The proposed optimized vocal model can generate glottal waveforms consistent with the actual voice source. And the physical parameters have significant differences among normal, polyps and paralysis vocal cords.