12th Speech in Noise Workshop, 9-10 January 2020, Toulouse, FR

Effects of asymmetric envelope compression on speech intelligibility and binaural unmasking

Emily Burg(a), Tanvi Thakkar, Sean Anderson
University of Wisconsin-Madison, USA

Matthew Winn
University of Minnesota, USA

Ruth Litovsky
University of Wisconsin-Madison, USA

(a) Presenting

The temporal envelope of a speech signal plays an instrumental role in speech understanding and perceptual segregation of speech and noise. Previous work has shown that compressing the envelope of vocoded speech (effectively reducing the dynamic range) can decrease speech understanding, but it is unknown whether dynamic range affects binaural integration and auditory source segregation. In order to investigate this, we examined the effect of reduced dynamic range on binaural unmasking in normal-hearing (NH) individuals. Specifically, we were interested in the degree of similarity in dynamic range across ears, because asymmetries in dynamic range are common in individuals with cochlear implants (CIs) and may contribute to their limited access to binaural benefits. We chose a binaural unmasking paradigm to assess binaural processing while avoiding effects that stem from simply attending to whichever ear has the better signal-to-noise ratio. The temporal envelope of the signal was compressed independently in one or both ears to simulate variations in dynamic range. Target stimuli consisted of female-talker IEEE sentences and masker stimuli consisted of female-talker AzBio sentences, all in English. Stimuli were processed using a 16-channel vocoder and presented over headphones at 65 dBA. In one condition, target speech was presented in quiet. In a second condition, target and masker speech were presented to one ear. In a third condition, a copy of the masker was also presented to the contralateral ear. When compared to performance in condition two, this third condition has been shown to elicit binaural unmasking in NH individuals, with improvements in speech reception thresholds ranging from 3-5 dB. We predicted that speech intelligibility in quiet and in noise would decrease monotonically with decreasing dynamic range. For bilateral conditions, we predicted that reducing dynamic range symmetrically in both ears would result in more unmasking than reducing it asymmetrically in one ear alone, because binaural integration requires similar representation and fidelity of signals across ears. Preliminary data show that speech intelligibility decreased with reduced dynamic range. Additionally, symmetrical reductions in dynamic range elicited more unmasking than asymmetrical reductions. This indicates that unmasking is not simply limited by the ear with the smallest dynamic range, and that asymmetries in dynamic range may be one factor contributing to the poor binaural processing demonstrated by individuals with bilateral CIs.

Last modified 2020-01-06 19:23:55