Analysis by synthesis of pathological voices using the Klatt synthesizer. Bangayan, P.; Long, C.; Alwan, A. A; Kreiman, J.; and Gerratt, B. R Speech Communication, 22(4):343-368.
doi  abstract   bibtex   
The ability to synthesize pathological voices may provide a tool for\textbackslashnthe development of a standard protocol for\textbackslashn\textbackslashnassessment of vocal quality. An analysis-by-synthesis approach using\textbackslashnthe Klatt formant synthesizer was applied to study 24\textbackslashn\textbackslashntokens of the vowel rar spoken by males and females with moderate-to-severe\textbackslashnvoice disorders. Both temporal and spectral\textbackslashn\textbackslashnfeatures of the natural waveforms were analyzed and the results were\textbackslashnused to guide synthesis. Perceptual evaluation\textbackslashn\textbackslashnindicated that about half the synthetic voices matched the natural\textbackslashnwaveforms they modeled in quality. The stimuli that\textbackslashn\textbackslashnreceived poor ratings reflected failures to model very unsteady or\textbackslashn��gargled�� voices or failures in synthesizing perfect copies\textbackslashn\textbackslashnof the natural spectra. Several modifications to the Klatt synthesizer\textbackslashnmay improve synthesis of pathological voices. These\textbackslashn\textbackslashnmodifications include providing jitter and shimmer parameters; updating\textbackslashnsynthesis parameters as a function of period, rather\textbackslashn\textbackslashnthan absolute time; modeling diplophonia with independent parameters\textbackslashnfor fundamental frequency and amplitude variations;\textbackslashn\textbackslashnproviding a parameter to increase low-frequency energy; and adding\textbackslashnmore pole-zero pairs.
@article{bangayan_analysis_1997,
	Author = {Bangayan, Philbert and Long, Christopher and Alwan, Abeer A and Kreiman, Jody and Gerratt, Bruce R},
	Date = {1997},
	Date-Modified = {2017-04-19 08:04:06 +0000},
	Doi = {10.1016/S0167-6393(97)00032-0},
	Issn = {01676393},
	Journal = {Speech Communication},
	Keywords = {clinical, clinical phonetics, Klatt, phonation, speech synthesis, speech technology},
	Number = {4},
	Pages = {343-368},
	Title = {Analysis by synthesis of pathological voices using the Klatt synthesizer},
	Volume = {22},
	Abstract = {The ability to synthesize pathological voices may provide a tool for{\textbackslash}nthe development of a standard protocol for{\textbackslash}n{\textbackslash}nassessment of vocal quality. An analysis-by-synthesis approach using{\textbackslash}nthe Klatt formant synthesizer was applied to study 24{\textbackslash}n{\textbackslash}ntokens of the vowel rar spoken by males and females with moderate-to-severe{\textbackslash}nvoice disorders. Both temporal and spectral{\textbackslash}n{\textbackslash}nfeatures of the natural waveforms were analyzed and the results were{\textbackslash}nused to guide synthesis. Perceptual evaluation{\textbackslash}n{\textbackslash}nindicated that about half the synthetic voices matched the natural{\textbackslash}nwaveforms they modeled in quality. The stimuli that{\textbackslash}n{\textbackslash}nreceived poor ratings reflected failures to model very unsteady or{\textbackslash}n��gargled�� voices or failures in synthesizing perfect copies{\textbackslash}n{\textbackslash}nof the natural spectra. Several modifications to the Klatt synthesizer{\textbackslash}nmay improve synthesis of pathological voices. These{\textbackslash}n{\textbackslash}nmodifications include providing jitter and shimmer parameters; updating{\textbackslash}nsynthesis parameters as a function of period, rather{\textbackslash}n{\textbackslash}nthan absolute time; modeling diplophonia with independent parameters{\textbackslash}nfor fundamental frequency and amplitude variations;{\textbackslash}n{\textbackslash}nproviding a parameter to increase low-frequency energy; and adding{\textbackslash}nmore pole-zero pairs.},
	Bdsk-Url-1 = {http://dx.doi.org/10.1016/S0167-6393(97)00032-0}}
Downloads: 0