Acoustical characterization and perceptual comparison of four types of 3Dprinted vocal tract models for the German and Japanese vowels /a,e,i,o,u/

Abstract:
Four types of 3D-printed vocal tract models for the German and Japanese vowels /a,e,i,o,u/ with different degrees of geometrical simplification and acoustical tuning were acoustically characterized and perceptually compared in order to facilitate an informed choice among such model types for education and research. They were acoustically characterized by means of their measured transfer functions for the closed-glottis condition and their resonance frequencies, and perceptually compared by 23 German and 23 Japanese subjects in a listening experiment with stimuli that were created by the voiced excitation of the models using a reed source. The results suggest that German and Japanese listeners generally prefer models of their respective native language, and among those, models with a lower degree of abstraction and a higher degree of acoustical tuning, with exceptions that can be related to other model properties like the wall thickness and the minimum cross-sectional area.