Understanding human voice disorders

Diseases of the human voice production apparatus represent far-reaching cuts in the quality of life of the affected individuals. As the human organ for voice production, the glottis is of special interest in investigating these disorders. The technical progress of the last years and decades makes it possible to study these diseases systematically through computer models. In the present master thesis, four different possible functional-based disease patterns are analyzed. These are (i) four different variants of initial glottal opening, (ii) symmetric and asymmetric vocal fold movements of the same frequency, (iii) vocal fold movements with unequal frequencies of the two vocal folds (periodic or aperiodic), and (iv) three different subglottal pressures. By combining these disease parameters, 48 configurations are formed, for which a hybrid aeroacoustic simulation model is used to calculate the resulting acoustic pressure. The hybrid aeroacoustic approach consists of an incompressible flow simulation from which the source term for the acoustic simulation arises. A filtering algorithm was developed to pre-process the aeroacoustic source terms, which allows attenuation of physically implausible impulsive structures in the source term signal. Using the resulting acoustic signals at the microphone point as well as the globally averaged source term signals, features from speech signal processing are evaluated (Harmonics-to-Noise-Ratio, Cepstral Peak Prominence, Hammarberg Index, Alpha Ratio, and Spectral Slope). A comprehensive cause-and-effect analysis is performed, for which the cause is represented by the simulation configuration parameters and the effects are the different evaluated features. The results exhibit general trends as well as specific deviations thereof. A future extension of the present model to organic-based voice disorders promises enhanced clinical applicability of the model.