VOWEL IDENTIFICATION IN MIXED-SPEAKER SILENT-CENTER SYLLABLES

Strange [J. Acoust. Soc. Am. 85, 2135-2153 (1989b)] has demonstrated t hat there is sufficient information in the onsets and offsets of sylla bles, spoken in sentence context, to provide accurate identification o f the vowel in the syllable. Verbrugge and Rakerd [Language Speech 29, 39-57 (1986)] and Andruski and Nearey [J. Acoust. Sec. Am, 98, 390-49 0 (1992)] have shown that such information is present in citation-form syllables even when the syllables begin with one speaker and end with another. These studies of ''hybrid syllables,'' however, reported rel atively high error rates. In a perceptual experiment using /dVd/ sylla bles spoken in sentence context by a male and a female speaker, relati vely low error rates were obtained for both ''silent-center'' syllable s and ''hybrid silent-center'' syllables. It was concluded that the in formation specified over syllable onsets and offsets together for iden tification of vowels is speaker independent and that it was sufficient in most cases to specify the vowel. The acoustic patterns, represente d as functions of log(F2/F1) over time, revealed potentially useful dy namic acoustic characteristics of coarticulated vowels.