A model of vowel production under positive pressure breathing

* Now at: 20/20 Speech Ltd., Malvern Hills Science Park, Malvern, Worcs. WR14 3SZ. [email protected] Abstract Agile combat aircraft currently in development using speech recognition as part of the cockpit interface will also use positive pressure breathing (PPB) to enable the pilot to remain conscious and function at high G levels. Recogniser performance is seriously affected by PPB even at moderate G levels when the intra-oral pressure is only 10-20 mmHg. This paper describes work which extends the standard n-tube model of vowel production to include the effects of intra-oral pressure. The ultimate aim of this work is to improve speech recogniser performance under these highly stressed conditions. An 8-tube Dynamic Regions and Modes model is used for this work, with the assumption that all regions of the vocal tract wall have the same compliance. A side branch was added to the model at the junction of the larynx tube and the lower pharynx to simulate the oesophagus. The model shows that as the intra-oral pressure is increased, the vowel space in the F1/F2 plane shrinks towards the region of F1 = 400 Hz, F2 = 1200 Hz. Measurements made on real speech recorded from six subjects at pressures up to 30 mmHg show a similar trend, although the reduction in the range of F2 is less than that indicated by the model. Variation in compliance of different regions of the vocal tract wall seems likely to be the main reason for the differences between the predictions of the model and the measured values.