Physical modeling of the singing voice

The singing voice is one of the most versatile musical instruments. In spite of the fact that singing can be learned by almost everybody, the physics of the singing voice are quite complex. This work presents a time-domain model for synthesis of sung vowels. The symmetric vocal fold model is based upon Titze’s 16-mass model [1] but takes into account jet generation and vortex shedding. Two vocal tract models have been implemented: a reflection-type line analog and a fast multiconvolution algorithm. The combined vocal fold and vocal tract model includes aspiration noise generation and can be adjusted during phonation for vibrato and transitions, e.g. between voice registers. The results are compared to measured mouth impedances and pressure transfer functions that were obtained using novel measurement techniques. The model can be used as a tool for visual and aural verification of algorithms for voice or speech synthesis or as a didactical tool for therapeutic professionals as well as for interested patients.