Predicting the speech intelligibility under fluctuating background conditions

Researcher(s): K.S. Rhebergen, PhD.
Supervisor(s): N.J. Versfeld, PhD., prof. W.A. Dreschler, PhD.

Background:
One of the main limitations who contribute to the experience of having a large handicap for hearing impaired is the bad speech intelligibility in background noise.
More than half a century ago researchers started to model this problem and now the SII model (Speech Intelligibility Index; American National Standard ANSI S3.5-1997 “Methods for Calculation of the Speech Intelligibility Index”) is largely used to predict the speech intelligibility of normal and hearing impaired listeners in a given background noise.
However, the SII model does not take into account any fluctuations in the masking noise; thus the model is only valid for steady state noises. Since most real-life noises do exhibit strong intensity variations over time or frequency (interfering speakers, music, traffic, etc), there is a great interest in an adequate procedure to predict speech intelligibility in interfering real life noises. Normal hearing listeners are somehow able to use the relative silent periods (“holes”) in a fluctuating background noise and therefore they are much better in hearing in fluctuating noises as in continuous steady state noises with the same level. This difference in signal to noise ratio can be as large as 6 dB to 20 dB. A hearing impaired listener is unfortunately unable to use these “holes” in the background noise. It is not difficult to understand that this handicap is precisely in social events (reception, party, meeting, etc) a large burden which can lead to a social isolation.

Aim:
The aim of this project is to model the speech intelligibility in fluctuating (real life) background noise. In the first instant the possibility of an extension of the SII is examined.

Method:
To validate the extension to the SII many speech intelligibility tests will be conducted in a diversity of background noises. The aim will be at real life and artificial noises. At first, all tests will be with normal hearing listeners. In a later phase of this study data of hearing impaired listeners will also be examined.
It is known that the SII model is unable to use non-linear enhanced signals (such as compression which is commonly used in hearing aids). This limitation of the SII will be examined in the second phase of this project.

Results:
The present study was started in the middle of 2002. During this project there will be a cooperation with the department of otolaryngology, experimental audiology of the VU Medical Center (VUmc). This cooperation implies as yet to a regular research meeting to discuss the experimental setup and the evaluation of the results.
In 2003 en 2004 many speech intelligibility tests have been performed by normal hearing listeners. With aid of many data from the literature an extension to the SII model is made (Rhebergen & Versfeld, 2005). On the base of this publication, the Acoustical Society of America (ASA) Working Group S3-79 asked the authors to take part of this workgroup, to prepare a modification to the existing ANSI S3.5 standard (2007) in which the extension for speech in fluctuating noise will be included.

Future research:
The SII extension (Rhebergen & Versfeld, 2005) will be validated in 2005 with data of normal hearing listeners collected in our lab in 2003 & 2004. Parallel to this the SII-model will be validated and extended with aid of existing data of normal and hearing impaired listeners from earlier studies of our lab and from the VUmc.
Furthermore, an SII extension will be prepared for the extension of the ANSI S3.5 standard.
The intention is to finish this project with taking a PhD degree.


Home Onderwijs Wetenschap Research (EN)
General Otology Prevention of hearing loss Auditory profile Sensorineural hearing loss and loudness perception Predicting the speech intelligibility under fluctuating background conditions Auditory filter measurements Auditory profile: spectral and temporal resolution (HEARCOM, WP2) Consonant confusions for filtered and consonant deteriorated nonsense VCV syllables (HEARCOM, WP2) Auditory localisation (HEARCOM, WP2) Dead Regions The perception of amplitude modulated signals Rehabilitation Hearing at work Publications Presentations	Predicting the speech intelligibility under fluctuating background conditions Researcher(s): K.S. Rhebergen, PhD. Supervisor(s): N.J. Versfeld, PhD., prof. W.A. Dreschler, PhD. Background: One of the main limitations who contribute to the experience of having a large handicap for hearing impaired is the bad speech intelligibility in background noise. More than half a century ago researchers started to model this problem and now the SII model (Speech Intelligibility Index; American National Standard ANSI S3.5-1997 “Methods for Calculation of the Speech Intelligibility Index”) is largely used to predict the speech intelligibility of normal and hearing impaired listeners in a given background noise. However, the SII model does not take into account any fluctuations in the masking noise; thus the model is only valid for steady state noises. Since most real-life noises do exhibit strong intensity variations over time or frequency (interfering speakers, music, traffic, etc), there is a great interest in an adequate procedure to predict speech intelligibility in interfering real life noises. Normal hearing listeners are somehow able to use the relative silent periods (“holes”) in a fluctuating background noise and therefore they are much better in hearing in fluctuating noises as in continuous steady state noises with the same level. This difference in signal to noise ratio can be as large as 6 dB to 20 dB. A hearing impaired listener is unfortunately unable to use these “holes” in the background noise. It is not difficult to understand that this handicap is precisely in social events (reception, party, meeting, etc) a large burden which can lead to a social isolation. Aim: The aim of this project is to model the speech intelligibility in fluctuating (real life) background noise. In the first instant the possibility of an extension of the SII is examined. Method: To validate the extension to the SII many speech intelligibility tests will be conducted in a diversity of background noises. The aim will be at real life and artificial noises. At first, all tests will be with normal hearing listeners. In a later phase of this study data of hearing impaired listeners will also be examined. It is known that the SII model is unable to use non-linear enhanced signals (such as compression which is commonly used in hearing aids). This limitation of the SII will be examined in the second phase of this project. Results: The present study was started in the middle of 2002. During this project there will be a cooperation with the department of otolaryngology, experimental audiology of the VU Medical Center (VUmc). This cooperation implies as yet to a regular research meeting to discuss the experimental setup and the evaluation of the results. In 2003 en 2004 many speech intelligibility tests have been performed by normal hearing listeners. With aid of many data from the literature an extension to the SII model is made (Rhebergen & Versfeld, 2005). On the base of this publication, the Acoustical Society of America (ASA) Working Group S3-79 asked the authors to take part of this workgroup, to prepare a modification to the existing ANSI S3.5 standard (2007) in which the extension for speech in fluctuating noise will be included. Future research: The SII extension (Rhebergen & Versfeld, 2005) will be validated in 2005 with data of normal hearing listeners collected in our lab in 2003 & 2004. Parallel to this the SII-model will be validated and extended with aid of existing data of normal and hearing impaired listeners from earlier studies of our lab and from the VUmc. Furthermore, an SII extension will be prepared for the extension of the ANSI S3.5 standard. The intention is to finish this project with taking a PhD degree.