PERCEPTUAL AND COMPUTATIONAL SEPARATION OF SIMULTANEOUS VOWELS - CUESARISING FROM LOW-FREQUENCY BEATING

Identification of simultaneous speech sounds, such as pairs of steady- state vowels (double vowels), is more accurate when there is a differe nce in fundamental frequency (F-0). Accuracy of identification for dou ble vowels increases with increasing F-0 difference (Delta F-0) asympt oting above 1 semitone. The experiment described here attempts to dist inguish two mechanisms underlying this effect: first, perceptual separ ation by grouping together harmonic components of a common F-0; and, s econd, exploitation of the fluctuations in the spectral envelope of th e composite stimulus that result from beating between unresolved compo nents. The beating is mainly caused by interactions between correspond ing harmonics of the two vowels with a small Delta F-0. Identification accuracy for normal, harmonically excited double vowels was compared with that for double vowels composed from the same components, but who se constituent vowels were excited by a mixture of the two harmonic se ries. These double vowels were designed to produce similar beating pat terns to the normal double vowels. Both harmonically and inharmonicall y excited constituents improved identification with increasing Delta F -0, but the increase was larger for harmonically excited vowels. A com putational model based upon psychophysical measurements of auditory fr equency and temporal resolution correctly predicted an increase in acc uracy of identification with increasing Delta F-0, which was attributa ble to beating. The results are interpreted in terms of a spectral cha nge cue in the identification of double vowels with Delta F-0's which complements grouping by F-0, and which plays a dominant role for Delta F-0's smaller than 1 semitone.