Ingrid Noordhoek: thesis announcement for the auditory list (Dan Ellis )

Subject: Ingrid Noordhoek: thesis announcement for the auditory list
From:    Dan Ellis  <dpwe(at)ICSI.BERKELEY.EDU>
Date:    Tue, 8 Aug 2000 12:14:14 PDT

Dear List - I was sent the enclosed announcement of a new thesis to circulate on the list. Best, -- DAn Ellis Auditory List Administrator ------- Forwarded Message From: "Noordhoek, Ingrid" <Noordhoek(at)TPD.TNO.NL> Subject: thesis announcement for the auditory list Date: Tue, 8 Aug 2000 11:44:41 +0200 [...] Recently, I have finished my thesis, entitled "Intelligibility of narrow-band speech and its relation to auditory functions in hearing-impaired listeners". I would like to inform the aditory list that my thesis may be downloaded from my web site at I am also pleased to mail paper copies to those who are interested as long as they are available. Below you will find the abstract of my dissertation. Sincerely, Ingrid Noordhoek. TNO Institute of Applied Physics Acoustics Division Phone +31 15 2692412 Fax +31 15 2692111 email: noordhoek(at) Listeners with sensorineural hearing losses often experience difficulties in understanding speech, even when the speech is presented well above their hearing thresholds. In this thesis, the origin of these difficulties was investigated by examining the relations between speech intelligibility and basic properties of the auditory system. To enhance the chances of obtaining clear correlations, the investigations were all performed in a limited frequency region around 1 kHz. A novel intelligibility test (the SRBT test) to measure intelligibility of speech bandpass filtered with a fixed center frequency of 1 kHz was developed in chapter 2. In this test, the minimum speech bandwidth required for a 50% intelligibility score is determined (speech-reception bandwidth threshold or SRBT). The narrowband speech is presented in complementary bandstop-filtered noise to ensure that the speech is only audible within the desired frequency band. The bandwidth of the speech signal is varied in an adaptive up-down procedure using a step size of a factor of 1.37 for the bandwidth (in Hz). On average, the SRBT of normal-hearing listeners is 1.4 octave (600-1600 Hz) under optimal conditions, i.e., when the entire speech dynamic range is above the hearing threshold, but not so loud that audibility is affected by excessive upward spread of masking. In chapters 3 and 4, the performance of, in total, 22 normal-hearing listeners and 53 hearing-impaired listeners was measured using the SRBT test, as well as using more common broadband SRT (speech-reception threshold) tests, namely the SRT in quiet and the SRT in noise. For most hearing-impaired listeners, the scores on these intelligibility tests deviated from those of the normal-hearing listeners. The speech intelligibility index (SII) model was applied to separate the origin of an elevated SRT or a broader-than-normal SRBT into (1) reduced audibility, and (2) suprathreshold deficits. Reduced audibility includes all those factors that would affect audibility in normal-hearing listeners under the same conditions when the hearing threshold of the hearing-impaired listener is simulated by presenting a masking noise. The effect of suprathreshold deficits on speech perception is quantified by the additional amount of speech information needed for an intelligibility score of 50% (as compared to normal-hearing listeners operating in threshold-simulating noise). For short sentences, normal-hearing listeners require an SII of about 0.3 for a 50% intelligibility score. This means that normal-hearing listeners need 30% of the speech information to understand 50% of the sentences correctly. When the SII value that a hearing-impaired listener requires for a 50% intelligibility score is normal, it is assumed that a possible deviant SRT or SRBT is due only to inaudibility of a part of the speech spectrum. On the other hand, a higher-than-normal SII value indicates that speech intelligibility was affected by suprathreshold deficits. The results of chapters 3 and 4 show that the presence and size of the effects of suprathreshold deficits on speech perception depend on the type of intelligibility test. The SRT test in quiet shows the smallest sensitivity to suprathreshold deficits in speech perception, while the SRBT test shows the largest sensitivity. Only a weak relation is observed between suprathreshold deficits and hearing loss. This shows that it is not possible to predict whether an individual listener suffers from a suprathreshold deficit from only a consideration of the size of the sensorineural hearing loss. In chapter 4, the relations between suprathreshold speech-perception deficits and basic auditory functions were investigated. The auditory-function tests included detection efficiency, temporal resolution (i.e., forward and backward masking), spectral resolution (i.e., upward and downward spread of masking), temporal and spectral integration, and discrimination of intensity, frequency, rhythm, and spectro-temporal shape. All auditory functions were measured at or around 1 kHz. Because several of these auditory functions were correlated, the thresholds on the auditory-function tests were subjected to a principal-components analysis. This resulted in three uncorrelated "auditory factors" (i.e., linear combinations of the auditory functions). The first factor is related to temporal resolution and frequency discrimination, the second factor is associated with spectral resolution, and the third factor is associated with detection efficiency, and temporal and spectral integration. Multiple regression was used to predict suprathreshold speech perception from the auditory factors for each of the intelligibility tests. The total variance accounted for by the auditory factors was largest for the results on the SRBT test (62%). This was in accordance with our expectations, because all auditory functions were measured around 1 kHz which is also the center frequency used in the SRBT test. The first and second auditory factors were most closely related to suprathreshold deficits in speech perception. This leads to the main conclusion of this thesis: the suprathreshold deficits that affect speech perception are (1) reduced temporal resolution, (2) reduced frequency discrimination, and (3) reduced spectral resolution. ------- End of Forwarded Message

This message came from the mail archive
maintained by:
DAn Ellis <>
Electrical Engineering Dept., Columbia University