Logo image
Back
What the /f/? We’re not done with fricatives yet. Integrating across time and frequency bands
Abstract

What the /f/? We’re not done with fricatives yet. Integrating across time and frequency bands

Bob McMurray, Marcus Galle, Ashley Farris-Trimble and Michael Seedorff
The Journal of the Acoustical Society of America, Vol.144(3 Supplement), pp.1716-1716
09/2018
DOI: 10.1121/1.5067610

View Online

Abstract

Fricatives represent an extreme version of lack of invariance. This illustrated by Jongman, Wayland and Wong’s (2000) singularly comprehensive look at the acoustics of a set of speech sounds. They show that fricatives require dozens of cues; and each cue is affected by multiple context factors. Later work with this corpus solved these problems with simple models. But we’re not done yet. I present new studies that raise two additional problems highlighted by fricatives. First, listeners must integrate cues over long timescales: frication cues may arrive several hundred milliseconds before the vocoid. I present several experiments that used eye-tracking to determine when cues are used. Unlike, prior results with stops and vowels, they show evidence for an encapsulated memory buffer preceding lexical access—lexical decisions are delayed until both cues arrive. Second, for fricatives, listeners must integrate information across distinct frequency bands. This was highlighted in a study of hybrid cochlear implant users who integrate low frequency residual acoustic hearing with high frequency electric hearing. Unexpectedly, hybrid listeners performed worse on fricatives than those with only electric hearing. Thus, the problem of combining information across frequency bands may be suboptimally solved in acoustic + electric CI users, suggesting new principles of cue-integration.

Details

Metrics

15 Record Views
Logo image