Date Published: December 12, 2017
Publisher: American Speech-Language-Hearing Association
Author(s): Kasey M. Jakien, Sean D. Kampel, Meghan M. Stansell, Frederick J. Gallun.
To evaluate the test–retest reliability of a headphone-based spatial release from a masking task with two maskers (referred to here as the SR2) and to describe its relationship to the same test done over loudspeakers in an anechoic chamber (the SR2A). We explore what thresholds tell us about certain populations (such as older individuals or individuals with hearing impairment) and discuss how the SR2 might be useful in the clinic.
Fifty-four participants completed speech intelligibility tests in which a target phrase and two masking phrases from the Coordinate Response Measure corpus (Bolia, Nelson, Ericson, & Simpson, 2000) were presented either via earphones using a virtual spatial array or via loudspeakers in an anechoic chamber. For the SR2, the target sentence was always at 0° azimuth angle, and the maskers were either colocated at 0° or positioned at ± 45°. For the SR2A, the target was located at 0°, and the maskers were colocated or located at ± 15°, ± 30°, ± 45°, ± 90°, or ± 135°. Spatial release from masking was determined as the difference between thresholds in the colocated condition and each spatially separated condition. All participants completed the SR2 at least twice, and 29 of the individuals who completed the SR2 at least twice also participated in the SR2A. In a second experiment, 40 participants completed the SR2 8 times, and the changes in performance were evaluated as a function of test repetition.
Mean thresholds were slightly better on the SR2 after the first repetition but were consistent across 8 subsequent testing sessions. Performance was consistent for the SR2A, regardless of the number of times testing was repeated. The SR2, which simulates 45° separations of target and maskers, produced spatially separated thresholds that were similar to thresholds obtained with 30° of separation in the anechoic chamber. Over headphones and in the anechoic chamber, pure-tone average was a strong predictor of spatial release, whereas age only reached significance for colocated conditions.
The SR2 is a reliable and effective method of testing spatial release from masking, suitable for screening abnormal listening abilities and for tracking rehabilitation over time. Future work should focus on developing and validating rapid, automated testing to identify the ability of listeners to benefit from high-frequency amplification, smaller spatial separations, and larger spectral differences among talkers.
In all tests, participants were asked to identify the target CRM phrase spoken by a male talker (the color–number combination after the call sign “Charlie”, always at 0° azimuth angle) in the presence of two masking sentences that were also spoken by male talkers. The masker and target talkers varied from trial to trial. Three talkers total were used. The fourth male talker in the corpus was not used due to his slower rate of speaking. Maskers could be either colocated with the target at 0° or symmetrically offset to the left and right by one of a variety of spatial separations. Responses were obtained using a monitor located in front of the participant showing an interface displaying all color and number options. Participants initiated each track and were given feedback (“Correct” or “Incorrect”) after each answer.
One of the main differences between the SR2 and the SR2A observed in Experiment 1 was the 2-dB difference in separated thresholds observed in the SR2 between runs 1 and 2. One possibility is that it reflects a genuine difference in repeatability between the two measures. Another possibility is that 27 of the 29 participants had already run the SR2 before being tested on the SR2A. To try to distinguish between these possibilities, in the second experiment, data from eight runs of the SR2 were examined for 40 participants, 16 of whom were also participants in Experiment 1.
The two experiments described above demonstrate that the SR2 is a reliable method of evaluating speech-on-speech masking and SRM. Experiment 1 revealed strong relationships between headphone and anechoic chamber testing. Experiment 2 revealed small differences across repeated testing for the same 40 participants.