Acoustic Analysis And Synthesis Of Pathological Voice

. Objective methods for evaluation of pathological voices have long been sought for both clinical use and basic research; classification of types and severity of voice defects has been largely accomplished by trained clinicians using subjective rating systems. However, these systems are very dependent on the evaluators. Automatic modeling and parameterization of pathological vowel samples/(?)/could provide a useful advance in this effort. Towards this goal, this work included the following sections:Pathological voices and normal voices were collected in clinical bases such as hospitals. A computer program was developed for automatic analyzing voice parameters based on linear prediction and inverse filter. Analysis-by-Perception experiments were designed for correlations between subject rates via GRBAS and analytical data with rank-sum test, and parameterized voice assessment functions was acquired by discriminant analysis experiments with Mahalanobis distance.A pathological voice source model was modeled in line with vocal parameters with significant difference based on LF model, for pathological voice analysis and synthesis. Further, a pathological voice synthesizer was developed for voice acoustical research, and could release the non-periodic components of sustained vowels/(?)/, which are prominent features of pathological voices, modeled as three phenomena:frequency jitter, amplitude shimmer, and aspiration noise.Synthesis-by-Analysis experiments were designed for verifying consistency of control parameters in synthesizer and analytical parameters, and Synthesis-by-Perception experiments were designed for verifying consistency of control parameters in synthesizer and subject rates to synthetically voice and value of functions.Pathological voice analysis and synthesis are the basis and precondition of Analysis-by-Perception-by-Synthesis methods of voice research. This study acquired parameterized voice assessment functions and a pathological voice source model by Analysis-by-Perception experiments for pathological voices. This study also verified value of functions and validity of the pathological voice source model by Synthesis-by-Perception experiments for pathological voices. Moreover, a pathological voice synthesizer was developed, which improved the research foundation of this field in China, enriched and developed the methods and tools of voice disorder research.