GLOTTAL MODELS FOR DIGITAL SPEECH PROCESSING - A HISTORICAL SURVEY AND NEW RESULTS

Glottal modeling has been an important topic of research in digital sp eech processing for many years. The ability to accurately model the gl ottal excitation is important for applications as varied as acoustic a nd articulatory speech synthesis, speech coding, and speech analysis. Many glottal models that differ in form and complexity have been sugge sted over the years. Possible models range from simple parametric mode ls of the glottal volume velocity or the glottal flow derivative that assume linear separability of the glottal source and the vocal tract t o more complex parametric function and mechanical models that allow fo r interaction between the glottal source and the vocal tract to very c omplex models that are based directly on the physiological properties of the glottis. This paper will provide a historical survey of glottal models, discussing their form and complexity along with the applicati ons for which each is appropriate. This paper will also present a disc ussion of the problem of modeling the glottal excitation of different styles of speech, a topic that is important for applications such as n atural, high-quality speech synthesis. A glottal model that is capable of modeling eleven commonly encountered styles of speech will be pres ented. (C) 1995 Academic Press, Inc.