Title: Asymmetrical Signal Degradation Causes Difficulty in Understanding Speech in Spatial Attention Tasks
Authors: Daniel Eisenberg and Matthew J. Goupell
Speaker: Daniel Eisenberg
Date: October 23, 2013; 12-1 PM
Room: LeFrak Hall Room 2208
Abstract:
Normal-hearing (NH) listeners can nearly perfectly recall words in an attended ear while ignoring the other ear when speech is presented to both ears simultaneously over headphones. However, recent experiments have shown that this does not necessarily occur in people with asymmetrical hearing. People with asymmetrical hearing loss, bilateral cochlear implants, or single-sided deafness very likely have a relatively “good” ear (i.e., high-quality speech signals) and “bad” ear (i.e., low-quality speech signals). We simulated different levels of signal degradation by independently varying the signal quality in each ear, as would occur in users of one or two cochlear implants, in order to understand the effect of having a good and bad ear in spatial attention tasks.
Participants were presented with two different sentences, one in each ear simultaneously, over headphones. In study 1, the quality of the signal was varied independently between the ears in each trial. The stimuli were either unprocessed speech, or noise vocoded speech with 4, 8, or 16 information channels. In experiment 2, the quality and the target-to-masker ratio (TMR) was varied. In experiment 3, the stimuli were either unprocessed speech, unshifted 8-channel sine vocoded speech, or 8-channel sine vocoded speech shifted either 3 or 6 mm. In all three studies, participants were told which ear to listen to (selective attention task), and were then asked to report the words they heard in that target ear. In experiment 1 there was an additional task in which participants were not told which ear the target was until after the stimuli were presented (divided attention task).
Results showed that if presented with a low quality target sentence in one ear and a higher quality interferer in the other ear, participants had a hard time ignoring the interferer and focusing on the target. In experiment 2, thresholds decreased as the number of channels in the target ear increased. However, the number of channels in the interfering ear had no effect on performance. In experiment 3, a definite learning trend was seen for the understanding of spectrally shifted speech. These data demonstrate that it is more difficult to attend to a poor-quality speech signal when trying to ignore a high-quality speech signal. This may adversely affect the advantage of spatial listening in certain spatial configurations for people with asymmetrical hearing, i.e., those with good and bad ears.
