by A.A. Kressner, A. Westermann, J. Buchholz and C.J. Rozell
Abstract:
Speech perception for most cochlear implant recipients is very limited in challenging listening environments. It has been shown that intelligibility outcomes in these situations can be improved with noise reduction or channel selection based on the ideal binary mask processing strategy. In realistic scenarios where prior information is unavailable however, the ideal binary mask must be estimated, and these estimations will inevitably contain errors. Although the effect of both unstructured and structured binary mask errors has been investigated with normal-hearing listeners, it has not been investigated with cochlear implant recipients. This study assesses speech recognition of binary-masked speech with cochlear implant recipients using binary masks with model-generated errors. The results demonstrate that clustering of mask errors substantially decreases the tolerance of errors, that incorrectly removing target-dominated regions can be more detrimental to intelligibility than incorrectly adding interferer-dominated regions if the errors are clustered, and that the individual tolerances of the different types of errors decreases when both are present. These trends follow those of normal-hearing listeners. It is clear from this study that it is essential to consider structure in the analysis of noise reduction and channel selection strategies that attempt to approximate the ideal binary mask processing strategy.
Reference:
Cochlear implant speech intelligibility outcomes with structured and unstructured binary mask errorsA.A. Kressner, A. Westermann, J. Buchholz and C.J. Rozell. Journal of the Acoustical Society of America, 139(2), pp. 800–810, February 2016.
Bibtex Entry:
@Article{kressner.15,
author = {Kressner, A.A. and Westermann, A. and Buchholz, J. and Rozell, C.J.},
title = {Cochlear implant speech intelligibility outcomes with structured and unstructured binary mask errors},
abstract = {Speech perception for most cochlear implant recipients is very limited in challenging listening environments. It has been shown that intelligibility outcomes in these situations can be improved with noise reduction or channel selection based on the ideal binary mask processing strategy. In realistic scenarios where prior information is unavailable however, the ideal binary mask must be estimated, and these estimations will inevitably contain errors. Although the effect of both unstructured and structured binary mask errors has been investigated with normal-hearing listeners, it has not been investigated with cochlear implant recipients. This study assesses speech recognition of binary-masked speech with cochlear implant recipients using binary masks with model-generated errors. The results demonstrate that clustering of mask errors substantially decreases the tolerance of errors, that incorrectly removing target-dominated regions can be more detrimental to intelligibility than incorrectly adding interferer-dominated regions if the errors are clustered, and that the individual tolerances of the different types of errors decreases when both are present. These trends follow those of normal-hearing listeners. It is clear from this study that it is essential to consider structure in the analysis of noise reduction and channel selection strategies that attempt to approximate the ideal binary mask processing strategy.},
year = 2016,
journal = {Journal of the Acoustical Society of America},
month = feb,
volume = {139},
number = {2},
pages = {800--810},
url = {http://scitation.aip.org/content/asa/journal/jasa/139/2/10.1121/1.4941567}
}