• Normal hearing listeners show perceptual dereverberation in speech recognition tasks. • Including preceding reverberated stimuli improves recognition of the final stimulus. • Full digit triplet presentation improves last digit recognition by 26.5 RAU or 1.5 dB SNR79. • Recognition improvement is due to dereverberation, not adaptation to noise. • CI users also benefit, but other effects may contribute to the improvement. This study investigated perceptual dereverberation in normal-hearing (NH) adults and cochlear implant (CI) users. Perceptual dereverberation refers to a reduction in the detrimental effect of reverberation on target speech recognition when preceding speech that causes the reverberation is included in the presentation. Recognition was compared between conditions with identical, equally reverberated target stimuli, with the preceding reverberant stimuli either included or omitted. The first experiment demonstrated perceptual dereverberation in NH listeners using a within-subject repeated measures design. Speech recognition was assessed using digit triplet and consonant-vowel-consonant word tasks in simulated reverberation and noise conditions. Digit triplets and word pairs were reverberated and presented either in their entirety or with only the final word or digit(s) included. Recognition of the last digit improved by 26.5 RAU, or 1.5 dB SNR 79 when the entire digit triplet was presented compared to presentation of only the last digit. For words, the improvement was 9.0 RAU. To explore whether the observed improvement was different to adaptation to noise, a second experiment was conducted. Recognition improvements were larger in reverberation compared to masking noise. In a third experiment, CI users showed that final digit recognition improved by 18.6 RAU when the digit triplet was presented as a whole. However, it cannot be ruled out that other effects, such as the abrupt start of the presentations, may have contributed to this finding in CI users. Our results suggest that normal-hearing listeners are able to utilize perceptual dereverberation to enhance speech recognition in reverberant environments.
Holtrop et al. (Sun,) studied this question.