Multiband compression limiting for hearing-impaired listeners.

Four multiband compression limiters and two linear amplification systems were compared in terms of the intelligibility of consonant-vowel-consonant (CVC) nonsense syllables for two hearing-impaired listeners over a 30 dB range of input levels. Each system incorporated one of two frequency-gain characteristics and one of three limiting characteristics (no limiting, moderate limiting, or severe limiting). The subjects were instructed to choose overall listening levels that would permit speech spanning the range of input levels to be as intelligible as possible and comfortable for long-term listening. Relative to linear amplification, the overall gain selected by the subjects increased by roughly 5 and 11 dB for the moderate and severe limiter, respectively. With linear amplification, the maximum score, 82 percent correct, was obtained at the highest input level and scores fell roughly 34 percentage points as input level was reduced. With compression limiting, although the maximum scores, 81 percent and 79 percent correct, were obtained at lower input levels, performance was comparable to that with linear amplification. Also, scores spanned a range of only 22 and 9 percentage points across the range of input levels with the moderate and severe limiter, respectively. This benefit was due to the improved scores provided by compression limiting at the low input levels. However, this advantage was offset somewhat by the disadvantage provided by compression at high input levels relative to linear amplification. Error analysis indicated that the spectral degradations introduced by independent compression of 16 frequency bands may have caused the reduced intelligibility at higher input levels.