Evaluation of a noise-reduction algorithm that targets non-speech transient sounds

Using a linear hearing aid simulation in the laboratory, 21 hearing-impaired listeners evaluated a transient noise-reduction algorithm in response to speech in quiet and speech in five typical background noises that had varying degrees of transience. The noise-reduction algorithm was tested when disabled and at minimum, medium, and maximum strengths, although only the disabled condition and medium strength were tested with all stimuli. A paired comparison test and absolute rating of annoyance test demonstrated a benefit from the noise-reduction algorithm on all strengths when the noise consisted of repeated transients and was presented at a somewhat high input level (door slamming and nails being hammered). However, for more continuous, fluctuating, and softer noises, such as paper rustling and cutlery clattering, maximum noise reduction was required to see significant benefit. For the loud, repeated transient noises, the noise-reduction algorithm reduced the annoyance of noise to below the percept of normal-hearing listeners. The noise-reduction algorithm had no effect on speech presented in quiet or in a more stationary party noise. Neither was there a significant effect of the noise reduction on horizontal localization performance using nail hammering as the stimulus or on speech recognition in paper rustling and cutlery clattering noises. These results were obtained with a computer simulation of the hearing aids, which had a long processing delay, and ears that were fully occluded by unvented earmolds. It is our understanding that wearable devices incorporating this noise-reduction algorithm have a much shorter processing delay, such that the amplification can be combined with vents and open fittings. Still, the transient noise reduction will, of course, apply only to the amplified, and not the vent-transmitted, portion of sound. Also, the results were obtained with the noise-reduction algorithm implemented in a linear scheme. Nevertheless, the effect of the algorithm on transient sounds achieved in this study should be maintained when it is used with modern non-linear devices, as long as any gain-changing processing with slower time constants than the transient noise-reduction algorithm occurs in the device after the transient noise-reduction processing. Therefore, we conclude that overall the noise-reduction algorithm for transient non-speech sounds offers benefit to the hearing aid user in terms of comfort without any noticeable side effects.