Syllable-boundary magnitudes from jaw movement patterns

Abstract

The C/D model [O. Fujimura, J. Acoust. Soc. Jpn. (E) 13, 39–48 (1992)] describes rhythmic patterns of utterances by amplitude-controlled pulse trains, each pulse representing a syllable or a boundary. The syllable magnitude (pulse height) controls mandibular movement. Assuming a direct proportional relation between the two variables, the syllable magnitude for each occurrence of five, nine, and Pine under various prominence conditions was inferred using Erickson’s Pine Street dialogue data [Erickson, Fujimura, and Pardo, Lang. Speech 41, 395–413 (1998)], which contained repeated corrections of the street address. A computational procedure has been devised to process semi-automatically microbeam data produced by two male and two female speakers. The timing of each syllable pulse was determined using the iceberg method [O. Fujimura, J. Acoust. Soc. Am. 99]. Syllable durations were determined in proportion to the syllable magnitudes. Boundary magnitudes were represented by the gaps among syllable durations. The syllable and boundary durations thus determined reasonably agreed with acoustic durations. [This work was supported in part by NSF Grant Nos. SBR-951199B, SBR-9809046, and BCS-9977018 (O. Fujimura) and by ATR/MIC, Japan.]