Voice source properties of the speech code

Abstract

A challenge for speech research and technology is to advance the knowledge-based approach. A neglected aspect of the speech code is the dynamics of the voice source and how it integrates in an overall production model. In the past, research into the human voice source has mostly been concerned with individual voice qualities. Knowledge of segmentally and prosodically induced variations and of waveform finestructure is now developing but the picture is still incomplete. An elaborated model of speech production integrating supraglottal, glottal, and laryngeal articulations and pulmonary activity is needed. The status of the art with a preliminary overview of contextual and prosodic factors that influence the voice source within a sentence frame will be discussed. One important factor is the degree of articulatory emphasis and thus of VT–source interaction as found in hypo- versus hyperspeech. Novel data obtained in Stockholm on subglottal pressure variations measured directly through a tracheal puncturing probe have contributed insight into intensity-F0 relations at covarying subglottal pressures. These data apply to global intensity declination within a phrase and to local regions of prominent F0 peaks where the normal rising relation of intensity with increasing F0 is canceled or reversed.