病理嗓音的声学分析和合成

发布时间：2018-07-31 05:28

【摘要】：对病理嗓音客观评估方法的研究,在言语嗓音的临床应用研究和基础研究领域中都具有重要的社会意义和理论意义。目前许多临床治疗师已经用主观评级的方法对病理嗓音进行了分类,但是这种主观评级方式对评价者非常依赖。理想的病理嗓音的评价应该是通过模型参数的匹配来对取样样本进行自动分析,从而得到更为客观的鉴别力。因此,本研究利用稳态元音/?/作为声学材料进行病理嗓音的声学分析和合成,通过分析合成感知实验来提出和验证嗓音音质评价的多参数变量函数方程组。围绕此研究目标,本研究工作主要以下几部分：在病理嗓音的音质特征研究基础上,利用线性预测分析和逆滤波技术,开发病理嗓音的声学分析软件,提取一系列描述嗓音音质特征的声学参数；并针对GRBAS进行主观感知将其等级分类。采用秩和检验方法验证主观感知结果与声学分析数据间的相关性,从而获取与感知评价一致的敏感声学参数；然后在此基础上利用马氏距离进行判别分析,以求解多参数变量的函数方程组,实现嗓音音质的客观评估。在病理嗓音敏感参数的研究基础上,以LF模型为基础进行病理嗓音源建模,开发病理嗓音信号合成器,实现病理嗓音在声学上的非周期性特征(包括基频微扰、振幅微扰和送气噪音),进而合成病理嗓音。在此基础上进行合成感知实验,对病理嗓音信号合成器的控制参数与合成嗓音的主观评价结果做相关性分析,从而进一步来验证评价函数的准确性和病理嗓音源建构的有效性。病理嗓音的分析和合成,是嗓音研究方法中分析合成感知法的基础和前提。本研究在国内外嗓音研究的基础上,通过病理嗓音分析感知的实验研究,获得了多维参数的嗓音评估函数和病理嗓音源模型；并通过病理嗓音的合成感知的实验研究来验证其函数的准确性和病理嗓音源模型的有效性。同时研究和开发了病理嗓音分析和合成软件,改善了国内在病理嗓音信号合成器方面的研究基础的不足,丰富和发展了嗓音障碍的研究方法和工具。
[Abstract]:The objective evaluation method of pathological voice is of great social and theoretical significance in the field of clinical application and basic research of speech voice. At present, many clinical therapists have classified the pathological voice by subjective rating method, but this subjective rating method is highly dependent on the evaluator. The ideal evaluation of pathological voice should be to analyze the sample automatically by matching the model parameters, so as to obtain more objective discriminant ability. Therefore, in this study, the steady-state vowel / is used as acoustic material for acoustic analysis and synthesis of pathological voice, and the multiparameter function equations for voice quality evaluation are proposed and verified by analyzing and synthesizing perception experiments. The main work of this study is as follows: based on the study of the acoustic quality of pathological voice, the acoustic analysis software of pathological voice is developed by using linear predictive analysis and inverse filtering technology. A series of acoustic parameters are extracted to describe the sound quality of the voice, and the GRBAS is classified by subjective perception. The correlation between subjective perception results and acoustic analysis data is verified by rank sum test, and sensitive acoustic parameters consistent with perceptual evaluation are obtained, and then Markov distance is used for discriminant analysis. The objective evaluation of voice quality is realized by solving the functional equations of multi-parameter variables. Based on the research of pathological voice sensitive parameters and LF model, the pathological voice signal synthesizer is developed to realize the acoustical characteristics of pathological voice (including fundamental frequency perturbation). Amplitude perturbation and exhalation noise), thus synthesizing the pathological voice. On the basis of these experiments, the correlation between the control parameters of the pathological voice signal synthesizer and the subjective evaluation results of the synthetic voice is analyzed, so as to verify the accuracy of the evaluation function and the validity of the construction of the pathological voice source. The analysis and synthesis of pathological voice is the basis and premise of synthetic perception in voice research. On the basis of the research of voice at home and abroad, through the experimental research of pathological voice analysis and perception, the multi-dimensional parameter voice evaluation function and pathological voice source model are obtained. The accuracy of the function and the validity of the pathological voice source model are verified by the experimental study of synthetic perception of pathological voice. At the same time, the software of pathological voice analysis and synthesis has been studied and developed, which has improved the deficiency of the research foundation of pathological voice signal synthesizer in China, and enriched and developed the research methods and tools of voice disorders.
【学位授予单位】：华东师范大学
【学位级别】：博士
【学位授予年份】：2011
【分类号】：R767.92

【参考文献】