Spatial Hearing in Speech Mixtures

Project Summary/Abstract
Communicating in environments containing multiple talkers (restaurants, parties, etc.) is the biggest challenge
faced by listeners with hearing loss. Previous work has shown that this difficulty stems from both monaural and
binaural deficits. Current hearing aids do not fully compensate for these deficits and thus do not offer adequate
improvements in multitalker environments. In some cases they even introduce additional distortions that may be
particularly detrimental in such situations. The goal of the proposed project is to leverage our understanding of
these different issues to explore ways of improving speech perception for hearing-aid wearers in multitalker
environments. Aim 1 will focus on restoring audibility, which is the primary goal of a hearing aid. Since
amplification strategies are mostly optimized for restoring speech audibility in quiet, they may not be optimized
for speech presented in multitalker mixtures. In particular, the speech of interest in multitalker mixtures may only
be available in rather sparse regions of time and frequency (or “glimpses”). The experiments proposed in Aim 1
will compare the ability of different amplification strategies to restore the audibility and intelligibility of glimpsed
speech. This work will demonstrate the impact that even small variations in audibility can have on real-world
speech understanding, and reveal potentially important differences between current amplification strategies. Aim
2 is motivated by the fact that some listeners with hearing loss also demonstrate reduced sensitivity to binaural
cues, which weakens the perception of location and may hinder selective attention. The experiments in this aim
will investigate speech enhancement strategies that are designed to enhance the representation of speech in
noise to improve intelligibility. Because they operate by altering the speech envelope, a secondary benefit may
be to increase the salience of binaural cues carried in the envelope. Enhancement will be applied to speech
stimuli across a range of spatial tasks, to identify conditions under which a binaural benefit is observed. This
work will open up a new avenue for improving the representation of complex acoustic scenes in hearing-aid
wearers. Finally, while appropriate amplification and binaural enhancement have the potential to improve speech
perception in multitalker mixtures, it is vitally important that hearing devices do not offset these benefits by
introducing distortions that have the opposite effect. Aim 3 will focus on one such distortion that has been
identified. While empirical data is limited, there are many indications that hearing aids increase the tendency for
external sounds to be perceived close to or inside the head. This disorienting effect may have particularly serious
consequences in multitalker mixtures, where distinct spatial locations are critical for segregating competing
voices. The experiments proposed in Aim 3 will systematically examine the effect of hearing aids on sound
externalization, clarify the source(s) of this effect, and determine the impact on speech intelligibility. This work
will highlight the critical importance of maintaining natural sound images in hearing aids for optimizing speech
intelligibility in real-world listening situations.

Grant Number: 5R01DC015760-10
NIH Institute/Center: NIH
Principal Investigator: Virginia Best