Analysis by synthesis of glottal airflow

Modelli ng of the glottal airflow has been performed with two different methods. One is with a frequency domain analog synthesizer, FLEA. The special aim of that study by Båvegård and Fant (1994a, 1994b) was to quantify and verify the interaction ripple on the flow, which is caused by the oral and tracheal loads when glottis is regarded as an oscill ator. In this case the glottal area variation with time is described in terms of the LF model of Fant, Lilj encrants and Lin (1985). This model can be characterized as a specific class of shapes that can be quantified with as littl e as three parameters. It can mostly be controlled to generate shapes having both temporal and spectral properties that are very similar to those of natural glottal flow functions. However, apart from the good match between model and data that can be achieved, it is still only a description of the end result delivered by the generation process. Some special features like the mentioned interaction with external mechanisms can be studied, but it does not in general explain how various features in the waveforms actually originated.