Toward formant synthesis with articulatory controls

This paper describes some recent advances in. formant synthesis of speech. Formant synthesis is based on an acoustic representation of speech-sound production in terms of sources of sound and filtering of these sources by the vocal tract. An extension of this source-filter model introduces a set of higher-level control parameters that specify how the subglottal pressure, the laryngeal configuration and state, and the vocal-tract configuration change with time during an utterance. A set of mapping relations derives the lower-level acoustic parameters from the higher-level quasi-articulatory parameters. Synthesis of the segmental, syllabic, and prosodic aspects of an utterance with this more compact set of higher-level parameters is illustrated with examples. Some advantages of this approach to formant synthesis are described.