Vowel and sibilant sound production in noise

Abstract

The present study investigated speech production in noise and whether speakers modulate the spectral contrasts of vowel and sibilant sounds depending on the frequency characteristics of the background noise signal. Twelve speakers participated in the present study. Speakers were presented with three-word sequences one word at a time on a computer monitor and, after a variable delay, produced the three words in the order they were presented. Word sequences were comprised of CVC words containing either or both of the vowels, /a/ and /ae/, and either or both of the sibilants, /s/ and /ʃ/, in word-initial position. During the presentation and production of each sequence, speakers were exposed to one of four possible noise conditions: (1) silence; (2) vowel masking; (3) sibilant masking; and (4) speech-shaped noise. The sibilant masker contained energy at frequencies associated with a speaker’s productions of /s/ and /ʃ/ and the vowel masker at frequencies associated with their productions of /a/ and /ae/. These masking signals were generated prior to the experiment based on practice productions of the stimulus words by each speaker. Spectral contrast distances are based on vowel formant frequencies and sibilant spectral moments (1–4). Noise condition effects on spectral contrast distances are being evaluated to determine whether speakers selectively alter their production of those sounds masked by background noise.