本文關(guān)鍵詞：病理喉聲源發(fā)聲系統(tǒng)模型的非線性分析及其優(yōu)化研究　出處：《蘇州大學(xué)》2015年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 聲帶疾病 非線性 聲帶建模 粒子群-擬牛頓

【摘要】：嗓音的好壞直接影響著語言的表達(dá),由于環(huán)境、職業(yè)等因素,聲帶疾病的患病率逐年增加,極大地影響了人們的健康和生活質(zhì)量。聲學(xué)檢測用于病理嗓音識別還處于理論研究階段,且對特征參數(shù)的研究仍存在不足,嗓音聲學(xué)特征優(yōu)化與實際聲帶生理結(jié)構(gòu)缺乏聯(lián)系。聲帶病變是導(dǎo)致嗓音異常的重要原因,近年來,關(guān)于聲帶的各類物理模型相繼被提出,各類病理聲帶發(fā)聲模型相關(guān)參數(shù)的研究為聲帶疾病分類提供了新的思路。本文針對聲帶疾病分類時參數(shù)選擇存在的局限性,從發(fā)聲系統(tǒng)建模的角度,提出對聲帶模型進(jìn)行連續(xù)的非線性分析,優(yōu)化模型參數(shù)匹配實際喉聲源,并分析息肉和麻痹聲帶模型的連續(xù)非線性參數(shù)以及物理參數(shù)的差異性。首先對正常和病理發(fā)聲系統(tǒng)振動進(jìn)行分析,考慮聲門下壓對發(fā)聲系統(tǒng)的影響,在不同聲門下壓條件下提取基頻、振幅相關(guān)參數(shù)和非線性動力學(xué)參數(shù),引入不同聲門下壓下的連續(xù)非線性參數(shù)并分析其差異性;其次采用優(yōu)化算法對聲帶模型進(jìn)行參數(shù)優(yōu)化,實現(xiàn)實際喉聲源的準(zhǔn)確仿真,使所建模型能夠準(zhǔn)確反映聲帶的生理結(jié)構(gòu),并對優(yōu)化得到的模型物理參數(shù)進(jìn)行差異性分析。具體研究內(nèi)容如下:(1)根據(jù)實際聲帶生理結(jié)構(gòu),分別建立正常和病理聲帶的發(fā)聲模型,并耦合聲門氣流產(chǎn)生喉聲源,求取喉聲源基頻和幅度相關(guān)參數(shù);引入龐加萊截面,分岔圖對模型振動進(jìn)行非線性分析;改變聲帶病理參數(shù)及聲門下壓,分析頻率參數(shù)和混沌參數(shù)李雅普諾夫指數(shù)的變化。仿真實驗結(jié)果表明:聲帶麻痹減小了嗓音發(fā)聲基頻,并且只在一定壓力范圍內(nèi)系統(tǒng)出現(xiàn)混沌,聲帶息肉引起的混沌則分布在整個發(fā)聲壓力范圍內(nèi)。(2)為了使聲帶模型參數(shù)和實際聲帶生理參數(shù)相匹配,采用模型結(jié)構(gòu)或參數(shù)的變化描述聲帶病變所引起的喉生理參數(shù)的改變,提出用粒子群-擬牛頓算法(PSO-QN)優(yōu)化聲帶生物力學(xué)模型參數(shù),匹配正常和病理喉聲源,得到反映實際喉參數(shù)的模型參數(shù),并對優(yōu)化得到的模型參數(shù)進(jìn)行差異性分析。本文引入病理喉聲源的聲帶物理模型,提出了連續(xù)聲門下壓下的最大李雅普諾夫指數(shù),并進(jìn)一步對聲帶模型參數(shù)進(jìn)行了優(yōu)化,結(jié)果表明,聲帶麻痹和聲帶息肉的最大李雅普諾夫指數(shù)隨聲門下壓變化的分布存在明顯差異性。本文提出的優(yōu)化聲帶模型可以產(chǎn)生與實際喉聲源一致的聲門波形,且優(yōu)化得到的正常、息肉和麻痹聲帶模型參數(shù)間存在顯著差異。
[Abstract]:The quality of voice directly affects the expression of language, due to the environment, occupation and other factors, the prevalence of vocal cord diseases is increasing year by year. It has greatly affected people's health and quality of life. Acoustic detection for pathological voice recognition is still in the theoretical research stage, and the study of characteristic parameters is still insufficient. There is no relationship between the optimization of voice acoustic characteristics and the actual physiological structure of vocal cords. Vocal cord lesion is an important cause of voice abnormalities. In recent years, various physical models of vocal cords have been proposed one after another. The research on the relevant parameters of various pathological vocal cord vocal models provides a new way of thinking for vocal cord disease classification. This paper aims at the limitations of vocal cord disease classification parameters selection from the point of view of vocal system modeling. A continuous nonlinear analysis of the vocal cord model is proposed to optimize the parameters of the model to match the actual laryngeal sound source. The continuous nonlinear parameters and physical parameters of polyp and paralytic vocal cord models are analyzed. Firstly, the vibration of normal and pathological vocal system is analyzed, and the influence of pressure under glottis on the vocal system is considered. The fundamental frequency, amplitude correlation parameters and nonlinear dynamic parameters are extracted under different glottic pressure conditions. The continuous nonlinear parameters under different pressure are introduced and their differences are analyzed. Secondly, the optimal algorithm is used to optimize the parameters of the vocal cord model to realize the accurate simulation of the actual laryngeal sound source, so that the established model can accurately reflect the physiological structure of the vocal cord. And the physical parameters of the optimized model were analyzed. The specific research contents are as follows: 1) according to the actual physiological structure of vocal cords, the vocal models of normal and pathological vocal cords are established respectively. The throat sound source is generated by coupled glottic airflow, and the fundamental frequency and amplitude parameters of the throat sound source are obtained. The nonlinear analysis of the vibration of the model is carried out by introducing the Poincare section and bifurcation diagram. The changes of Lyapunov exponent and frequency parameters were analyzed by changing the pathological parameters of vocal cord and glottic pressure. The simulation results show that vocal cord paralysis reduces the pitch frequency of voice. And only in a certain pressure range system chaos, vocal polyps caused chaos is distributed in the whole vocal pressure range. 2) in order to make the vocal cord model parameters and actual vocal physiological parameters match. The changes of model structure or parameters are used to describe the changes of laryngeal physiological parameters caused by vocal cord disease. PSO-QN algorithm is used to optimize the parameters of vocal cord biomechanical model. Matching the normal and pathological laryngeal sound sources, the model parameters reflecting the actual laryngeal parameters are obtained, and the differences of the optimized model parameters are analyzed. In this paper, the vocal cord physical model of the pathological laryngeal sound source is introduced. The maximum Lyapunov exponent under continuous threshold pressure is proposed, and the parameters of the vocal cord model are further optimized. The results show that. The distribution of the maximum Lyapunov exponent of the vocal cord paralysis and polyp with the change of the pressure under the glottis is obviously different. The optimized vocal cord model proposed in this paper can produce the glottic waveform consistent with the actual laryngeal vocal source. The parameters of normal, polyp and paralytic vocal cord models obtained by optimization are significantly different.
【學(xué)位授予單位】：蘇州大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2015
【分類號】：R767.4

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 尹輝;謝湘;匡鏡明;;基于聽覺模型與自適應(yīng)分?jǐn)?shù)階Fourier變換的聲學(xué)特征在語音識別中的應(yīng)用[J];聲學(xué)學(xué)報;2012年01期

2 周強;張曉俊;顧濟(jì)華;趙鶴鳴;朱俊杰;陶智;;嗓音多頻帶非線性分析的聲帶病變識別[J];聲學(xué)學(xué)報;2014年01期

，

本文編號：1413323