Exploring the methods of physics-based vocal synthesis. Weighing accuracy and performance in the formulation of an appropriate computational implementation.
Diet Strebe’s Robotic Mouth
Pink Trombone is an interactive vocal simulation that you can launch in your browser.
The Source-Filter Model
The model I’m going to be implementing in this project is the source-filter model. It
Avg chest voice phonation frequency = 80-400 for male and 200-500 for female
the Open Quotient is the fraction of the phonation period that the glottis opens and closes. During speech this value is usually between .4 and .8
Types of glottal phonation:
The waveform produced at the glottis can categorised into three discrete types. These are determined primarily on the size of the open quotient.
Breathy phonation is has a the shortest open quotient. As a result, its waveform is the most sinusoidal (resembling a sine wave) of the three.
Modal phonation is the normal way in which our glottis produces sound as we vocalise.
Laryngalised (Vocal Fry)
Turbulence noise theory using Reynold’s Number
High velocity glottal flow creates turbulence noise which is called aspiration noise if sourced from nearby. Aspiration is what gives the voice its breathy quality.
Vocal tract modelling
Average vocal tract diameter is around 2 cm
Wave-shape variable Rd used to transition between the modes of glottal phonation.