Theodore S. Bell
UCLA School of Medicine, Head & Neck Surgery, 31-24 Rehabilitation Ctr., Los Angeles, CA 90024-1794
Richard Wright
UCLA, Los Angeles, CA 90024-1543
Three experiments tested the effectiveness of adjusting the amplitude of a noise masker to selectively alter intelligibility of individual words within sentences. Stimulus sentences contained three key words; all were commonly used and phonetically unique. A speech-shaped noise was presented at 65 dB SPL for all experiments. The initial study tested 12 listeners under two conditions: the noise level remained flat, or was attenuated 6 dB under the middle key word. The speech was presented at 54 and 58 dB SPL. The middle key word improved by as much as 30% (p<0.001), while other key words were unaffected. In the next two experiments, the noise was either increased or decreased by 3 dB at a key word. In one, the manipulation was at the first key word, and in the other was at the last key word. In both experiments, the speech was presented to ten listeners at 56 dB SPL. The 3-dB increment in noise significantly decreased the word's intelligibility, and the 3-dB decrease increased the intelligibility (p<0.01) without affecting other key words in the sentence. This technique has application in reducing variability in sentence intelligibility and greatly reducing perceptual dynamic range. Implications for ``all-or-none'' contextual recognition in adverse conditions are discussed.