Towards sibilant physical speech screening using oral tract volume reconstruction: Some preliminary observations

The current paper considers physical speech screening by supplying air to a rigid replica based on oral tract volume reconstruction of a speaker uttering phoneme /s/ (Reynolds number about 5300). Radiated acoustic pressure spectra were measured for different flow conditions upstream from the reconstructed portion: Reynolds number (2800, 5300, 8900) and turbulence intensity (<4% and >4% by varying the method of air supply). Acoustic spectra obtained with the replica were compared to spectra of phoneme /s/ uttered by the same speaker at different loudness levels. It is found that noise emitted by the replica reproduces the spectral shape (in particular for frequencies up to 8 kHz) and the order of magnitude of spectral features (spectral slopes and dynamic amplitude) of phoneme /s/. Nevertheless, spectral differences (energy discrepancy, peak frequency, negative spectral slope) between phoneme /s/ and sound generated with the replica are observed as a function of Reynolds number as well as a function of upstream turbulence intensity. Therefore, current data suggest that in order to perform sibilant physical speech screening Reynolds number as well as upstream flow conditions need to be taken into account.