ASA 125th Meeting Ottawa 1993 May

5pSP3. Formant transitions and the identification of concurrent vowels

Peter Assmann

School of Human Devel., Univ. of Texas at Dallas, Box 830688, GR4.1, Richardson, TX 75083

Listeners exploit several sources of information to segregate the speech sounds produced by two competing talkers. When two synthetic vowels are presented simultaneously and monaurally, listeners can take advantage of a difference in fundamental frequency (f[sub 0]) to identify the vowels. This study compared the contributions of f[sub 0] differences and formant-frequency transitions to the identification of concurrent vowels. Accuracy improved when a two-semitone difference in f[sub 0] was introduced, compared to the case where both f[sub 0]'s were the same. In some conditions a small, independent improvement was observed when one of the vowels was preceded by a 100-ms gliding formant pattern. However, the presence of the glide did not help to identify the vowel which is preceded; its presence helped by making the competing vowel more identifiable. One explanation for improvement in the glide condition is a formant-tracking process which groups together the formants of each voice on the basis of good continuity. However, this account predicts improvement for both vowels which was not observed. An alternative account is provided by a model that (i) performs a series of analyses using a brief, sliding temporal window, and (ii) determines which region of the signal provides the strongest evidence of each vowel constituent. [Work supported by NIDCD.]