Masker location uncertainty reveals evidence for suppression of maskers in two-talker contexts

Abstract

In many natural settings, spatial release from masking aids speech intelligibility, especially when there are competing talkers. This paper describes a series of three experiments that investigate the role of prior knowledge of masker location on phoneme identification and spatial release from masking. In contrast to previous work, these experiments use initial stop-consonant identification as a test of target intelligibility to ensure that listeners had little time to switch the focus of spatial attention during the task. The first experiment shows that target phoneme identification was worse when a masker played from an unexpected location (increasing the consonant identification threshold by 2.6 dB) compared to when an energetically very similar and symmetrically located masker came from an expected location. In the second and third experiments, target phoneme identification was worse (increasing target threshold levels by 2.0 and 2.6 dB, respectively) when the target was played unexpectedly on the side from which the masker was expected compared to when the target came from an unexpected, symmetrical location in the hemifield opposite the expected location of the masker. These results support the idea that listeners modulate spatial attention by both focusing resources on the expected target location and withdrawing attentional resources from expected locations of interfering sources.

Effects of a consistent target or masker voice on target speech intelligibility in two- and three-talker mixtures

“…Thus, the constant masking voice did not appear to capture listeners' attention. Instead, these results are in line with reports of better perceptual segregation after priming with a multi-tone masker (Kidd et al, 2011) or with spatial location (Allen et al, 2011) or when the masker is a highly familiar voice (Johnsrude et al, 2013). Since the masker is outside of the focus of attention, such improved performance due to masker consistency cannot be due to template-matching (Bregman, 1990), but may be due to improved segregability of the two voices.…”

“…For instance, detection of a target narrowband tone-burst sequence embedded in multi-tone maskers was better when, on each trial, listeners were cued with the multi-tone maskers compared to when they heard a notched noise band (notch centered on the center frequency of the target) prior to the pattern-detection task (Kidd et al, 2011). In multitalker situations, presenting a masking speech signal at an expected versus an unexpected spatial location can improve target speech intelligibility (Allen et al, 2011), although this effect is not consistently observed (Jones and Litovsky, 2008). In a recent study examining the effects of voice familiarity on speech segregation using the CRM procedure (but not the CRM corpus of voices), listeners were significantly better at reporting coordinates from voices of strangers, ageand sex-matched to that of their spouse, when their spouse's voice was used as the masker in a two-talker mixture (Johnsrude et al, 2013), compared to when voices from other age-and sex-matched strangers were used as maskers.…”

Target identification using relative level in multi-talker listening

“…Both similarity and uncertainty contribute to interference between stimuli and produce informational masking (Watson, 2005). Uncertainty about a target's location, for example, has been found to have a detrimental impact on speech perception during multi-talker listening (Allen et al, 2011;Allen et al, 2009;Kidd et al, 2005) and reducing uncertainty about the target can provide large benefits to performance (Kidd et al, 2005) even when uncertainty about the masker location(s) remains (Jones and Litovsky, 2008).…”

High Oxygen Exchange to Music Indicates Auditory Distractibility in Acquired Brain Injury: An fNIRS Study with a Vector-Based Phase Analysis

“… 38 listening to two or more concurrent music streams (e.g., a melody with harmonic accompaniment or two melodies) activates different types of attention, such as selective alternating and divided attention. One of the musical features, instrumental timbres, provides a perceptual cue that aids segregating multiple streams and focusing on a relevant one, such as in speech perception in noise 39 – 42 . Moreover, multi-voice music listening tasks activate neural regions involved in different types of attention 43 – 46 .…”