J Am Acad Audiol 10 : 445-457 (1999)
Comparison of Performance with
Wide Dynamic Range Compression
and Linear Amplification
Anna C . S . Kam*'
Lena L . N . Wong*
Abstract
This study compared subject performance and preference using a compression-limiting hearing aid set to linear amplification (program 1) and wide dynamic range compression (WDRC,
program 2) . The frequency responses of the hearing aid were matched to a 65 dB SPL signal
and maximum output to a 90 dB SPL signal . Twenty subjects with moderate to moderately
severe sensorineural hearing loss were tested . Speech recognition scores and speech reception thresholds were obtained both in quiet and in noise . Subjective preference for WDRC
or linear amplification was measured via a paired-comparison procedure on "loudness appropriateness," "clarity," and "pleasantness" to continuous discourse presented in quiet and in
noise . Results suggested that WDRC yielded better speech intelligibility in quiet for low-level
signals and no difference in speech intelligibility in noise compared to linear amplification .
Subjects preferred WDRC for loudness to both high- and low-level signals and for pleasantness to high-level signals .
Key Words: Linear amplification, speech intelligibility, subjective preference, wide dynamic
range compression
Abbreviations : ANOVA= analysis of variance, AVC =automatic volume control, CL= compression limiting, CVR = consontant-vowel-ratio, HINT = Hearing In Noise Test, P = program,
REIG = real-ear insertion gain, REIR = real-ear insertion response, SNR = signal-to-noise
ratio, SRS = speech recognition score, SRT = speech reception threshold, WDRC = wide
dynamic range compression
onlinear amplification or compression
systems were originally incorporated
Nin hearing aid design to limit the maximum output of a hearing instrument to below
the wearer's loudness discomfort level. Before the
introduction of compression, peak clipping was
the most frequently used method of output limiting (Hawkins and Naidoo, 1993). However,
this method causes a considerable amount of distortion that is unpleasant to the listener and may
even reduce speech intelligibility (Dreschler,
1988a). To eliminate or minimize these undesirable effects, the use of compression limiting
was introduced as an alternative to peak clip*Department of Speech and Hearing Sciences, The
University of Hong Kong, Hong Kong ; tcurrently affiliated
with Phonak Hearing Centre, Hong Kong Limited
Reprint requests : Lena L. N. Wong, Department of
Speech and Hearing Sciences, Prince Philip Dental Hospital,
5/F, 34 Hospital Road, Hong Kong
ping. Except for people with very profound hearing loss, compression limiting is superior to
peak clipping both subjectively and objectively
(Gioannini and Franzen, 1978 ; Dreschler, 1988a ;
Hawkins and Naidoo, 1993 ; Dillon, 1996) .
The use of compression to limit hearing aid
output is just one of its applications in hearing aid
design . Based upon its function and characteristics, compression used in modern hearing aids can
be categorized into three main types: compression
limiting (CL), slow-acting automatic volume control (AVC), and wide dynamic range compression (WDRC) (Dillon, 1988 ; Hickson, 1994). The
characteristics of the three types of compression
systems are summarized in Table 1 .
CL is characterized by high compression
ratios, high compression thresholds, and short
time constants (attack time and release time).
This type of compression is designed to limit
the output below the listeners' tolerance level,
minimize temporal and spectral distortion of
445
Journal of the American Academy of Audiology/Volume 10, Number 8, September 1999
Table 1
Compression
System
CL
AVC
WDRC
Characteristics of Three Common Types of Compression System (after Fortune, 1996)
Compression
Threshold
(dB SPL)
>80
<65
<60
Compression
Ratio
>5
>5
<5
Attack
Release
(msec)
(msec)
<5
10-50
<5
50-100
150-2000
10-100
Time
Time
CL = compression limiting, AVC = automatic volume control, WDRC = wide dynamic range compression .
the acoustic input at low and intermediate levels by providing linear amplification to sounds
below the compression threshold, and minimize
distortion that may be generated by peak clipping (Dreschler et al, 1984 ; Boothroyd et al,
1988 ; Dillon, 1988 ; Preves, 1991).
Hearing aids with AVC typically have intermediate to high compression ratios, low compression thresholds, short attack time, and a very
long release time . Because of the long release
time, the output of the signal remains relatively
constant in the presence of input fluctuations.
As a consequence, the need to adjust the volume
control of the aid is reduced (Dillon, 1988 ; Hickson, 1994 ; Kuk, 1996).
WDRC is characterized by low compression
ratios, low compression thresholds, and short
time constants. Kuk (1996) and Dillon (1988) limited the definition of WDRC to a system with
short release time . This syllabic compression
system allows the compression aid to "follow" the
envelope fluctuation among syllables seen in
speech (Kuk, 1996). The rationale behind WDRC
is to match or "squeeze" the normal speech range
(from soft to loud speech) to the reduced dynamic
range of hearing-impaired people (Steinberg
and Gardner, 1937 ; Dillon, 1988). People with
sensorineural hearing loss typically have reduced
auditory dynamic range (the difference in dB
between detection threshold to threshold of discomfort) . In the unaided condition, low-intensity
sounds will be inaudible while high-intensity
sounds remain loud . Unlike linear amplification, which provides constant gains regardless
of the input sound levels, WDRC gives more
gain to low-intensity sounds and less gain (gain
reduction) to high-intensity sounds . As a result,
soft sounds should become audible while listening comfort is ensured even with loud sounds .
During the past 2 decades, numerous studies have been carried out to investigate the performance of different amplification systems .
Lines of research include comparison of performance between CL and peak clipping (e .g .,
Gioannini and Franzen, 1978 ; Dreschler, 1988a;
446
Hawkins and Naidoo, 1993), linear amplification
and AVC (e .g., Moore et al, 1991 ; Neuman et al,
1994), linear amplification and WDRC (e .g .,
Nabelek, 1983 ; Dreschler, 1988b ; Peterson et
al, 1990), and evaluation of multiband compression (e .g ., Plomp, 1988 ; Moore et al, 1992 ;
Hohmann and Kollmeier, 1995 ; Yund and Buckles, 1995) . However, the findings in most areas
are inconclusive .
Theoretical advantages of WDRC over linear amplification are well documented (e .g .,
Hickson, 1994 ; Dillon, 1996 ; Kuk, 1996). However, due to problems and differences in the
design of various empirical studies, results documenting the advantages are conflicting or
inconclusive . Verschuure et al (1993) investigated the effect of syllabic compression (i .e .,
WDRC) on speech intelligibility in 19 hearingimpaired listeners. Results obtained via a nonsense consonant-vowel-consonant syllable test
in quiet revealed better speech intelligibility
with WDRC . The study used experimental hearing aids that were specifically designed so that
many parameters of the instrument could be
manipulated. Most of the parameters studied
could not be altered in commercial hearing aids,
making comparisons to other studies very difficult, if not impossible . Accordingly, generalization of the benefits of compression from
experimental hearing aids is not readily applicable to commercially available instruments.
There are some tentative experimental findings for speech intelligibility improvement in
quiet with the use of WDRC instead of linear
amplification. For example, Dreschler (1988a)
compared the performance using a syllabic compressor with compression limiting and a linear
amplifier with peak clipping. Sixteen hearingimpaired subjects with hearing loss ranging
from mild to moderately severe participated in
a phoneme perception task carried out in quiet.
It was found that compression yielded significantly better phoneme identification scores than
linear amplification . Nabelek (1983) showed
that WDRC was superior to linear amplification
Comparison of WDRC and Linear Amplification/Kam and Wong
in quiet for 13 subjects with mild-to-severe sensorineural hearing loss . When speech-spectrumshaped noise was introduced, performance with
linear amplification was better than with WDRC .
An insignificant or negative effect of compression was reported in some studies . Dreschler
et al (1984) compared the performance among
five hearing aids (a linear aid, two input compressors, two output compressors) . Twelve hearing-impaired subjects were recruited . The speech
reception thresholds (SRTs) for sentence material were obtained in quiet and in noise . No significant difference in performance was observed
among these hearing aids . Using 16 subjects
with mild-to-severe sensorineural hearing loss,
Tyler and Kuk (1989) failed to find any significant improvement in consonant identification in
babble noise using a single-channel syllabic
compressor and its linear version .
Hickson et al (1995) evaluated the consonant
perception of 15 subjects with mild-to-moderate
sensorineural hearing loss using linear amplification and compression amplification with two
different compression ratios (1 .3 and 1 .8). No significant difference was found in the scores
obtained in a nonsense syllable test in quiet. In
the background of babble noise, consonant perception was significantly better with linear
amplification than with either form of compression . In this study, speech material was
first processed by the hearing instruments and
then recorded and played back via headphones
during testing. As the hearing aid was not individually fitted, the hearing-impaired subjects'
aided hearing ability was not optimized. This also
hindered the generalization of the research findings to real-life situations .
Different experimental tasks and conditions
were employed in the studies, making it difficult
to compare findings . For example, some studies
have used phonemes (e .g ., Dreschler, 1988b),
nonsense syllables (e .g ., Vershuure et al, 1993),
and sentences (e .g., Dreschler et al, 1984) as
the experimental speech material . When speech
intelligibility was measured in noise, some
experiments employed multitalker babble (e .g.,
Hickson et al, 1995), and some used speechspectrum-shaped noise (e .g ., Nabelek, 1983).
These differences in experimental tasks and
conditions also contributed to the inconsistency
of the experimental findings .
Experimental Rationale
Although the efficacy of WDRC has not been
fully validated, such a signal-processing method
has already been used in many commercially
available hearing aids for some time . Without
any well-validated evidence or rationale, what
can hearing aid dispensers do when they must
select the most appropriate hearing aid or cir-
cuit for their clients? Byrne (1996) suggested that
the hearing aid dispenser is losing control of
the fitting process . The most basic question, "Is
compression beneficial to hearing-impaired people?" needs to be answered .
An interesting follow-up question to the
above is "Is compression beneficial to Cantonesespeaking hearing-impaired people?" In comparison with English, which is an intonation
language (i .e ., language uses pitch variations
over phrases and sentences to distinguish meaning differences), Cantonese is a tone language,
which uses the pitch of individual syllables to
contrast meanings (Fromkin and Rodman,1988) .
According to Ladefoged (1993), English is a
stress-timed language, whereas Cantonese is a
syllable-timed language in that syllables tend to
recur at regular intervals of time and stress is
less important than other prosodic features such
as tone . These differences in suprasegmental
feature contribute to the difference in spectral
and temporal cues of the two languages . Accordingly, it is worthwhile to investigate the effect
of WDRC (i .e ., syllabic compression) on Cantonese perception. Previously, performance of
two-channel WDRC and one-channel linear
amplification on Cantonese perception had been
compared in a study done by Wong et al (1996) .
In the study, better speech intelligibility using
the two-channel aid was found in noise and in
quiet. To date, no information is available about
the performance of single-band WDRC in Cantonese perception .
The main aim of the present study was to
investigate the difference in performance with
single-channel WDRC and linear amplification
in (a) Cantonese speech intelligibility in quiet,
(b) Cantonese speech intelligibility in noise, and
(c) subjective preference . It was hoped that, with
the objective and subjective data, a better comparison of WDRC and linear amplification performance can be obtained .
METHOD
Subjects
Twenty subjects (9 male and 11 female)
ranging in age from 16 to 70 (mean = 44 .6, SD
= 17 .4) were selected from among experienced
hearing aid users who were visiting the
447
Journal of the American Academy of Audiology/Volume
10, Number 8, September 1999
Table 2 Pure-Tone Hearing Thresholds of the
Subject's Better Ears (Test Ears)
Subject
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Mean
SD
Hearing Level (dB HL) (re : ANSI S3.6-1989)
250
Hz
500
Hz
65 .00
65 .00
60 .00
65 .00
45 .00
55 .00
55 .00
60 .00
55 .00
45 .00
40 .00
65 .00
45 .00
55 .00
65 .00
60 .00
50 .00
45 .00
40 .00
65 .00
55 .00 65 .00 65 .00 70 .00 80 .00
55 .00 60 .00 70 .00 70 .00 80 .00
70 .00 60 .00 65 .00 60 .00 75 .00
65 .00 55 .00 55 .00 60 .00 70 .00
45 .00 55 .00 60 .00 60 .00 70 .00
50 .00 55 .00 50.00 55 .00 70 .00
55 .00 60 .00 50.00 50 .00 65 .00
60 .00 65 .00 65 .00 65 .00 75 .00
55 .00 55 .00 60 .00 60 .00 65 .00
50 .00 50 .00 60 .00 65 .00 70 .00
45 .00 50 .00 55 .00 55 .00 65 .00
60 .00 65 .00 60 .00 50 .00 65 .00
45 .00 45 .00 45 .00 50 .00 55 .00
60 .00 65 .00 60 .00 60 .00 70 .00
65 .00 65 .00 70 .00 70 .00 75 .00
55 .00 55 .00 55 .00 60 .00 70 .00
50 .00 60 .00 60 .00 65 .00 70 .00
40 .00 40 .00 45 .00 50 .00 65 .00
45 .00 45 .00 50 .00 50 .00 65 .00
65 .00 70 .00 70 .00 65 .00 75 .00
54 .50 57 .00 58 .50 59 .50 69 .75
8 .26
8 .01
7 .80
7 .05
5 .96
55.00
9 .03
1000
Hz
2000
Hz
4000-8000
Hz
Hz
Audiology Clinic of the Department of Speech
and Hearing Sciences at the University of Hong
Kong or the Phonak Hearing Centre Hong Kong
Limited. All had bilateral sensorineural hearing
loss . To be included in the study, the degree of
loss in the better ear (i .e ., the test ear) must be
in the moderate to moderately severe range (i .e .'
the pure-tone average for 0.5, 1, and 2 kHz lies
between 42 to 68 dB HL) and the configuration
of the loss must be flat (i .e ., the difference in
hearing level between 500 and 4 kHz was 15 dB
or less). Table 2 shows the pure-tone hearing
thresholds of the subjects' better ears (test ears).
Average pure-tone thresholds ranged from 40 to
67 dB HL (mean = 56 .7, SD = 7.3). All subjects
were experienced hearing aid users with hearing aid experience of 1 to 23 years (mean = 7.9,
SD = 6.6). Only Cantonese speakers with good
oral communication abilities were recruited.
The profile of the subjects is shown in Table 3.
Equipment and Procedure
All testing was carried out in a sound-treated
room at the Audiology Clinic of the Department
of Speech and Hearing Sciences at the University of Hong Kong . Each subject was tested for
about 2 hours in each session. The hearing aid
448
was programmed for the test ear (better ear)
before the subjects arrived . Hearing re-evaluation, hearing aid verification, speech intelligibility measurement both in quiet and in noise,
and subjective rating were done within the same
session.
Hearing Instrument
A Phonak behind-the-ear instrument, model
Piconet2 P2 AZ, was used as the experimental
hearing aid. Maximally, three hearing programs
could be set and the circuit type could be selected
independently in each program. The programs
set could be switched via a digital remote control . It had a Multi Dynamic Compression Control, which could be set to yield linear
amplification or WDRC . In this study, Program
1 (Pl) was set to linear amplification and Program 2 (P2) was set to WDRC . Both programs
exhibited compression limiting with adaptive
release time . The static and dynamic characteristics of the circuits are shown in Table 4.
Hearing Aid Programming
Before the subjects arrived, the target 2-cc
coupler gain based on the subjects' most recently
obtained audiogram was calculated using the
Table 3 Profile of Subjects (N = 20 )
Factor
Characteristics Data
Gender
Age
Type of hearing loss
Degree of hearing loss
9 male, 11 female
16-70 years old (mean = 44 .6)
20 sensorineural hearing loss
6 moderate (pure-tone average
of 0 .5, 1 kHz and 2 kHz
= 41-55 dB HL)
14 moderately severe (puretone average of 0 .5, 1 kHz, and
2 kHz = 56-70 dB HL)
20 flat (<15 dB difference
between 1 and 2 kHz)
3 congenital
1 head trauma
6 presbycusis
7 viral infection
3 unknown
10 monaural, 10 binaural
behind-the-ear users
1-23 (mean = 7 .9)
Audiogram
configuration
Cause of hearing loss
Hearing aid
experience
Years worn
hearing aids
Previous hearing aid
circuit type
2 WDRC with CL, 18 linear
with CL
WDRC = wide dynamic range compression, CL =
compression limiting .
Comparison of WDRC and Linear Amplification/Kam and Wong
Table 4
Circuit
Amplification Char acteristics of Different Circuits in the Experimental Hearing Aid
Compression Limiting
Compression Characteristics
CT
CR
AT
RT
Linear + CL
N/A
N/A
N/A
N/A
WDRC + CL
45
1 .1 :1 up
5
30
to 2.7 :1
1
CT
CR
80
8 :1
AT
5
RT
Adaptiv e
CT = compression threshold, CR = compression ratio, AT = attack time (msec), RT = release time (msec), CL = compression
limiting, WDRC = wide dynamic range compression, N/A = not applicable .
FIG6 (3 .0, Rev L) program. P1 of the instrument was set to linear amplification and P2 was
set to WDRC . The 2-cc coupler gain of P1 and
P2 was matched to the prescribed gain for 65 dB
SPL input. Because the Phonak aid is a singlechannel device, the compression ratio for WDRC
setting was set to the average value of the prescribed ratios for both the high- and low-frequency bands. The mean compression ratio used
for the group is 2.35 (SD = 0 .30) . FIG6 was
selected since it is designed for prescribing gains
and compression ratios for compression hearing
aids . The prescribed gain for 65 dB SPL input
was also used to prescribe gain for the linear
amplification to ensure similar response between
the two programs . As the OSPL 90 of P2 would
be set automatically according to the selected
compression ratio, no adjustment was made .
OSPL 90 of P1 was set to match that of P2 . The
volume control wheel and the on-off switch on
the instrument was deactivated.
Frequency Response Fine Adjustment
With the loudspeaker positioned 1 meter
from the subject at 0" azimuth, subjective feedback to 65 and 80 dB SPL (root mean square)
continuous discourse was collected. Subjects
were asked to rate the signals using a 7-point
scale: "cannot hear," "very soft," "soft," "comfortably loud," "loud," "very loud," and "intolerably loud ." The objective was to ensure that
conversational level (65 dB SPL) speech signals
were perceived as "comfortably loud" and high
input level (80 dB SPL) signals were not "intolerably loud ." It was found that no adjustment
was necessary for these subjects .
Hearing Aid Veriffcation
At the beginning of the session, a routine
re-evaluation of the subject's hearing was done .
The hearing thresholds of all subjects across all
of the tested octave frequencies were within ±5
dB of the previously obtained value. After confirmation of the subject's hearing status, hearing aid verification was done . Real-ear insertion
response (REIR) was measured using the composite noise signal from the Fonix 6500 real-ear
analyzer at three input levels : 50, 65, and 80 dB
SPL. These steps were performed to verify the
electroacoustic performance of the hearing aid
and to verify that the frequency response actually matched the target REIR . These input levels were also used to assess speech intelligibility.
Speech Material for Speech Intelligibility
Measure
The Monosyllabic Cantonese Word Lists
(Lau and So, 1988) and the Hearing In Noise Test
(HINT, Chinese version) (Nilsson et al, 1994 ;
Wong et al, in preparation) were used . The test
stimuli of both tests were recorded by a male
native Cantonese speaker. The speech stimuli
were recorded on a CD-ROM. The recordings
were played via a personal computer through a
GSI-16 audiometer to a loudspeaker located 1
meter from the subject at 0° azimuth.
The monosyllabic word lists were used to
measure word recognition ability. There were 10
lists of monosyllabic words and 1 list was presented in each testing condition to determine the
speech recognition score (SRS), which was
defined as the percent of syllables correctly
repeated .
The HINT (Chinese version) was used to
determine SRT, which was defined as the presentation level necessary for a listener to recognize the speech materials correctly 50 percent
of the time . Four 10-sentence lists, each with sentences of equal level of difficulty and phonemic
content, were used . One list was presented in
each testing condition.
449
Journal of the American Academy of Audiology/Volume 10, Number 8, September 1999
Speech Intelligibility in Quiet
One monosyllabic word list was presented
at each of three sound levels : 50, 65, and 80 dB
SPL. The presentation levels were selected to
evaluate the performance with sounds of low
input level just above the compression threshold of WDRC (50 dB SPL), high input level above
the compression threshold of compression limiting (80 dB SPL), and at everyday conversational speech level (65 dB SPL) . The subjects
were required to repeat aloud what they have
heard. SRSs were obtained for both hearing aid
programs .
For the HINT, one sentence list was presented with each hearing aid program. The subjects were instructed to listen and repeat aloud
whatever was heard or understood. An adaptive
up-down strategy described by Nilsson et al
(1994) was used to adjust the sentence presentation levels . The first sentence of a list was
presented below threshold and the level was
increased in 2-dB steps until the sentence was
repeated correctly. The subsequent sentences
were presented once each, with the presentation
level dependent upon the accuracy of the preceding response . Presentation levels were
decreased by 2 dB after a correct response and
raised by 2 dB after an incorrect response . SRT
using each program was estimated as the mean
presentation level calculated from the fifth to
tenth sentences in the list .
Speech Intelligibility in Noise
One monosyllabic word list was presented
in each testing condition. SRSs in noise with
noise fixed at 65 dB SPL and at -9, -6, -3, 0, +3,
+6, +9 dB SNRs were obtained. The levels were
chosen to survey a range of performance across
SNR. Afour-talker babble (two female, two male)
was used as competition signal during tests for
speech recognition in noise.
The sentence lists of HINT were presented
in a background of spectrally matched noise
fixed at 65 dB SPL. SRTs (in terms of SNRs) were
obtained by a similar procedure as that was
used in quiet test . For tests in noise, both signals and noise were presented from the same
loudspeaker.
For both tests in quiet and in noise, the
sequence of presentation of word lists and sentence lists and the order of the program being
evaluated were randomized . The sequence of
testing conditions or presentation levels was
also randomized to counterbalance any prac450
tice and fatigue effect . The subject was blinded
to the program in use and was not informed of
the difference in the programs before finishing
the experiment .
Subjective Preference Measure
Another set of 12 HINT sentences was used
for sound quality rating. One sentence was presented in each testing condition and 12 listening conditions were evaluated. Preference of
"loudness appropriateness," "sound clarity," and
"sound pleasantness" was rated for the programs in quiet at 50, 65, and 80 dB SPL and at
+6 dB SNR with the spectrally matched noise
fixed at 65 dB SPL. The three rating categories
were chosen as they were relatively more concrete and easier to define compared to other
dimensions, such as brightness, sharpness, spaciousness, or fullness, which had been commonly
measured in other studies (e .g., Balfour and
Hawkins, 1992 ; Lundberg et al, 1992). This
made the task simpler, especially for the older
subjects . The SNR of +6 dB was selected as the
test condition in noise because the speech intelligibility at this SNR had been rated as satisfactory to good by normal-hearing and
hearing-impaired subjects in Lazarus's (1985)
study (cited in Bachler and Vonlanthen, 1994).
Sound quality judgment on unintelligible speech
might be very difficult and unreliable .
A paired-comparison procedure was used
to obtain subject preferences for linear amplification (Pl) and WDRC (P2) . The same sentence
was presented in a listening condition while the
hearing aid was switched to either programs
via the remote control held by the experimenter.
In evaluating the preference for loudness appropriateness, the subject had to indicate which
presentation of the sentence sounded more suitably loud . For clarity, the subject had to indicate
which presentation of the sentence sounded
clearer, that is, from which of the two he/she
could extract the text more easily. Pleasantness
is independent of intelligibility. Hence, the subject had to indicate which presentation sounded
pleasanter regardless of the ease of understanding.
In each listening condition, the subject was
allowed to switch back and forth between presentations with P1 and P2 as many times as was
necessary before a decision was made . After the
subject had decided a preference for either program, he/she was asked to assign a strength to
that preference . That is, the subject had to indicate whether the preferred program was (a)
Comparison of WDRC and Linear Amplification/Kam and Wong
t 50 dB
WDRC
- 65 dB
WDRC
t 80 dB
WDRC
t 50 dB SPL
-
.n-- 50 dB
Lmea
o - 65 dBr
Linear
80 &B
L-
65 dB SPL
-+-80 dB SPL
0
750
500
1500
1000
2000
3000
6000
4000
FrequencYLHz;
5O
Figure l Mean REIGs for the WDRC program at three
input levels (N = 20).
much better, (b) moderately better, or (c) just
slightly better than the other program . The subject was also encouraged to explain the preferences . The subjects were blinded to the program
in use to minimize any subject bias . The sequence
of listening conditions and the order of programs being evaluated were randomized to
counter any task order effect .
RESULTS
Real-Ear Measurement
Mean real-ear insertion gains (REIGs) at
three tested input levels are shown in Figures
1 and 2 . REIGs using the two amplification
schemes are overlaid in Figure 3 . The gains
across all tested octave frequencies at 50 dB
SPL were significantly higher than that at 65 dB
SPL (p < .01) with the WDRC program, whereas
the gains at these two levels did not differ significantly with the linear program . Both programs provided significantly less gain at 80 dB
SPL than that at 65 dB SPL (p < .01) across all
tested frequencies .
When REIGs between the programs were
compared (see Fig . 3), significantly more gain
500
1000
1500
2000
3000
Frequency (FLz ;
4000
6000
Figure 3 Mean REIGs for the linear and WDRC programs at various input levels .
was provided by the WDRC program to 50 dB
SPL input at 250, 500, 1000, 1500, and 2000 Hz
(F = 7.94, p < .05) . There was no significant difference in REIG at the input level of 65 dB SPL.
At 80 dB SPL, the WDRC program provided
significantly less gain at 250 and 500 Hz (F =
7 .94, p < .05) than the linear program .
Speech Intelligibility In Quiet
As shown in Figure 4, the mean SRSs
obtained using the monosyllabic word test with
WDRC were better than those obtained with
linear amplification at all presentation levels in
quiet. A two-way analysis of variance (ANOVA)
with repeated measures on both factors (program
and presentation level) was performed to investigate the effects of type of amplification and test
condition. Results are presented in Table 5. Both
factors and their interaction were significant,
indicating that the effect of program is significantly different for distinct presentation levels .
As shown in Table 6, a matched-pair t-test
revealed a significant difference between programs only at the presentation level of 50 dB SPL
(t = 4.54, p < .01) .
Mean SRTs obtained using the HINT in
quiet with the two programs are shown in Figure 5 . Results of a matched-pair t-test revealed
a significantly better SRT with WDRC (t = 2 .36,
p < .05) .
t 50 dB SPL
--~ -- 65 dB SPL
--r--80 dB SPL
250
500
1000
1500
2000
Frequency (Hz',
3000
4000
6000
Figure 2 Mean REIGs for the linear program at three
input levels (N = 20).
Speech Intelligibility In Noise
Mean SRSs obtained using the monosyllabic word test in noise are shown in Figure 6.
Results of ANOVA of the data, as presented in
Table 7, revealed a significant effect of test condition (SNR) and a significant interaction of
program and test condition. This suggests that
the effect of program is significant in certain test
451
Journal of the American Academy of Audiology/Volume 10, Number 8,
September 1999
Table 5 Analysis of Variance for Speech Recognition Scores Obtained
in Quiet with Both Programs
Source
df Effect
MS Effect
df Error
MS Error
F
Program
Presentation level
Program x presentation level
1
2
2
1880 .21
9799 .38
473.96
19
38
38
136 .35
259 .02
51 .15
13 . 79"
37 . 83"
9 .27"
MS = mean square
.p < .01 .
conditions only. A matched-pair t-test was performed to investigate any significant difference
in performance in each test condition . Results
are summarized in Table 8. Better SRSs were
obtained with "RC at a SNR of +6 dB (t = 2 .49,
p < .05) and with linear amplification at SNRs
of 0and-6 dB (t=2 .87,p< .01 ;t=3 .24,p< .01) .
Mean SRTs obtained using the HINT in
noise are also shown in Figure 5. No significant
difference was found with matched-pair t-tests .
Subjective Preference
Subjective preference for the two different
hearing aid programs in the different test conditions is shown in Table 9. The strength of
preference for loudness, clarity, and pleasantness
under various test conditions is shown in Figures 7, 8, and 9, respectively. A sign test was carried out to investigate any significant difference
in preference . Significant subjective preference
was found for WDRC for loudness appropriateness to signals at 50 and 80 dB SPL and for
pleasantness to signals at 80 dB SPL (p < .05) .
No significant preference for clarity was found
in any test condition.
DISCUSSION
quiet. That is, WDRC yielded better word recognition in quiet, at least to low-level signals. One
possible reason for this finding is the better
audibility ensured by WDRC . WDRC provided
significantly more gain to input at 50 dB SPL
than did linear amplification . As the signal level
increased from 50 to 65 dB SPL, the gain provided by WDRC decreased to an amount approximately equivalent to that provided by linear
amplification, accounting for equivalent performance between programs . When the signals
were presented at a high level (80 dB SPL), the
gain provided by WDRC decreased further. For
linear amplification, the gain also decreased
due to activation of compression limiting, but to
a lesser extent than WDRC . Although signals
were louder, speech intelligibility did not improve
significantly with linear amplification. At such
a high presentation level, audibility is no longer
a dominating factor for speech intelligibility.
Other factors, such as distortion in the auditory
system or the hearing aid, contribute to the difficulties of speech perception experienced by
those with moderate or greater cochlear losses
(Moore, 1996).
For sentence material, the SRT obtained in
quiet with WDRC was significantly lower than
Speech Intelligibility in Quiet
The SRSs obtained in the monosyllabic word
test with WDRC were significantly better than
those with linear amplification at 50 dB SPL in
90
a
o.
bo
~ 50
------------ ------------------------°------------1----------------------- . ---
= Mean+SD
Mean-SD
°
Mean
40
,~ 30
- ---------
2 to
--------
"
o 10
a; o
.. .{. . .------~--- ..--;. . . .. . . . .
13
~ -10
a
~ -20
-------------------
---- - - ----- ' ----------------------- ..
WDRC In Quiet
Linear In Quiet
WDRC N Nom
T..--_-- . ;
Lirrear N Noise
Test Condition
Figure 4 Mean SRS (%) obtained in quiet with WDRC
and linear amplification.
452
Figure 5 Mean and SD of SRTs obtained in quiet and
in noise with both programs .
Comparison of "RC and Linear Amplification/Kam and Wong
Table 6 Matched-Pair t-test Results for Speech Recognition Scores Obtained in Quiet with Both
Programs
Program
Linear
WDRC
Presentation
Level (dB SPL)
Mean
SD
Mean
SD
df
t
50
69 .50
19 .53
54 .50
23 .11
19
4.54*
88 .75
91 .75
65
80
10 .87
9 .36
11 .01
6 .67
83 .50
90 .50
1 .64
0 .63
19
19
WDRC = wide dynamic range compression
*p < .01 ,
that obtained with linear amplification . That
is, subjects were better able to repeat sentences
at lower level in quiet with WDRC . The effect is
congruent with increased gain provided by
WDRC to low-level signals . Another possible
reason is the increase in consonant-vowel-ratio
(CVR) caused by WDRC, which ensures that
the more intense vowel sounds receive less gain
than the less intense consonant sounds (Hickson and Byrne, 1995 ; Kuk, 1996) . Speech intelligibility may be improved by increased CVR .
Increasing the consonant level should serve to
enhance the audibility of acoustic cues necessary
for perception . At the same time, decreasing the
vowel level should decrease the masking effects
of the stronger components of speech on the
weaker ones, thus also enhancing the audibility of consonant acoustic cues .
The findings in quiet were consistent with
those obtained in some other studies . For example, Dreschler (1988a) found better phoneme
identification in quiet with a compression aid
compared with a linear instrument in a group
of 16 hearing-impaired subjects . Using a modified rhyme test, Nabelek (1983) showed better
word recognition in quiet with WDRC in eight
OSRS osingWDRC
" SRS uvng Linear
--- Flt Ilne for WDRC
628
..
Fttlinefo,Linear
hearing-impaired subjects . Due to the large variation in methodology used in different studies,
it is impossible to make any fair comparison
between studies . However, as suggested by Kuk
(1996), one common conclusion could be drawn
from those studies that reported supportive evidence for WDRC : positive effect was observed in
tests when the stimulus was presented in quiet
at a fixed low level . Results from the present
study give further support to this conclusion.
Speech Intelligibility in Noise
The overall SRSs obtained in the monosyllabic word test and the SRT obtained using
HINT with WDRC were not significantly different from those obtained with linear amplification . In other words, WDRC did not provide
significant improvement in speech intelligibility in noise. This finding is not uncommon . Many
previous studies, which also employed a constant background noise in testing, reported similar results . For example, no significant
difference in SRT obtained in noise in 12 hearing-impaired listeners was reported by Dreschler
et al (1984) . Tyler and Kuk (1989) also failed to
Table 7 Analysis of Variance for Speech
Recognition Scores Obtained in Noise with
Both Programs
Source df Effect MS Effect
Program
SNR
q
6
3
0
-3
Signal-to-poise Ran. (SNR) (dB SPL I
-6
Figure 6 Mean speech SRS (%) obtained in noise with
WDRC and linear amplification .
Program
x SNR
df Error MS Error
F
6
29162.38
114
262 .98 110.89*
167 .76
1 .38
6
620 .36
114
152 .16
4 .08*
1
232.23
19
MS = mean square
SNR = signal-to-noise ratio.
*p < .01 .
453
Journal of the American Academy of Audiology/Volume 10, Number 8, September 1999
Table 8 Matched-Pair t-test Results for Speech Recognition Scores
Obtained in Noise with Both Programs
Program
WDRC
Linear
SNR (dB)
Mean
SD
Mean
SD
df
t
+9
+6
+3
0
-3
-6
79 .25
75 .50
68 .50
50 .00
39 .00
15 .00
13.79
15 .30
19 .27
19 .74
24 .47
16 .86
77 .50
66 .75
63 .50
62 .75
47 .00
23 .00
7 .00
12 .41
13 .79
20 .72
18 .95
21 .97
21 .67
12 .92
19
19
19
19
19
19
19
0 .50
2 .49*
0 .98
2 .88*
1 .51
3 .24**
0 .31
-9
7.50
13 .91
SNR = signal-to-noise ratio, WDRC = wide dynamic range compression .
*p< .05;**p< .01 .
demonstrate any significant improvement in
consonant recognition with WDRC over linear
amplification in a multitalker babble background
noise for 11 listeners.
Reasons for poor performance with WDRC
in noise were not well understood and have not
been verified. There are some proposed explanations . Tyler and Kuk (1989) suggested that the
temporal information contained in the speech signal may be disrupted by the dynamic amplification . Another possible reason is that
compression amplification may increase the
level of the background noise in the gaps of the
speech signal (if the noise is present at a lower
level than the speech) and hence cause masking
of the speech signal (Hickson et al, 1995).
The relationship between circuit type and
speech intelligibility requires further clarifica-
tion . Although WDRC did not improve speech
intelligibility in noise, it did not degrade the
performance in comparison to linear amplification. In the monosyllabic word test, at favorable
SNRs, the mean SRSs obtained with WDRC
were better than those with linear amplification.
In more adverse listening conditions, linear
amplification provided slightly better speech
intelligibility. For the sentence test, the SRT
obtained with WDRC was lower (i .e ., better
speech intelligibility) than that with linear
amplification, although the amount was not statistically significant .
The findings in the present study do not
support the notion that WDRC has a negative
effect on speech intelligibility in noise when
compared to linear amplification. However,
WDRC may be slightly more vulnerable to noise
Table 9 Subjective Preference for Program in Various Test Conditions (N = 20)
Quality
Signal Level
(dB SPL)
Number of Subjects
Preferring WDRC
Number of Subjects
Preferring
Linear Amplification
Loudness
80 in quiet
65 in quiet
50 in quiet
In noise at SNR = +6
15*
11
15*
11
5
g
5
80 in quiet
65 in quiet
50 in quiet
In noise at SNR = +6
7
10
11
9
13
10
9
11
15*
13
12
8
5
Clarity
Pleasantness
80 in quiet
65 in quiet
50 in quiet
In noise at SNR = +6
SNR = signal-to-noise ratio, WDRC = wide dynamic range compression.
*Sign test significant, p < .05.
454
9
7
8
12
Comparison of WDRC and Linear Amplification/Kam and Wong
IS
15
M WDRC much better
M WDRC much better
0 WDRC moderately better
E3 WDRC moderately better
p WDRC slightly better
p WDRC slightly better
M
Linear much better
Linear moderately better
0 Linear slightly better
0 Linear much better
~, t0
0 Linear moderately better
U
M Linear slightly better
w
O
N
E S
z
z
0
I
a-
80 dB SK In Quiet 65 dB SPL U Quiet
In Noise
50 dB SPL hi Quiet
Test Condition
Figure 7 Strength of subjective preference for "loudness"
in various test conditions .
than linear amplification . From Figure 6, it can
be seen that the regression line for WDRC is
steeper than for linear amplification . That is, a
shift from a higher SRS to a lower SRS takes a
smaller change in SNR for WDRC than for the
linear program . This finding could not be compared to related studies in the area as other
studies seldom varied the SNR . The implication of this observation is that when performance comparison between WDRC and linear
amplification is made, the use of a single SNR
is undesirable . A range of SNR should be
included to give a fair evaluation of both systems .
This issue had been discussed in Yund and Buckles's (1995) study, which evaluated the perfor-
IS
M
WDRC much better
WDRC moderately better
F71 WDRC slightly better
O, Linear much better
Linear moderately better
Linear slightly better
I
80 dB SPL In Quiet
65 dB SPL In Quiet
50 dB SPL In Quiet
In Noise
Test Condition
Figure S Strength of subjective preference for clarity
in various test conditions .
80 dB SPL In Quiet
65 dB SPL In Quiet
50 dB SPL In Quiet
InNoise
Test Condition
Figure 9 Strength of subjective preference for pleasantness in various test conditions .
mance of multichannel systems instead of single-channel compression.
Subjective Preference
Significant subjective preference was found
for WDRC for loudness appropriateness to signals at 50 and 80 dB SPL and for pleasantness
to signals at 80 dB SPL. The explanations for
these findings are straightforward . At 50 dB
SPL, the gain provided by WDRC was significantly higher than that provided by linear program . Better audibility gained more votes for
better loudness . In fact, most subjects complained
that the sound with the linear program was too
soft . Better audibility also resulted in significantly better speech intelligibility to 50 dB SPL
signals. At 80 dB SPL, significantly more gain
was provided by linear amplification. However,
the extra gain seemed to be too much for most
subjects in this study Although the stimuli at 80
dB SPL was not so pleasant, subjects' word recognition scores, compared to WDRC condition,
were not degraded . It seems that although the
signal was not pleasant, it was not distorted
enough to reduce speech intelligibility.
Interestingly, subjects did not show a significant preference for clarity in all test conditions. One possible reason is that once the stimuli
were audible or clear enough using either program, it was difficult for the subjects to tell
which program sounded clearer . Preference of
clarity does not seem to be affected by preference
of pleasantness .
Alarge intersubject variability was observed
in the preference judgment . From the strength
455
Journal of the American Academy of Audiology/ Volume
10, Number 8, September 1999
of preference shown in Figures 7, 8, and 9, it can
be seen that in one condition, different subjects
preferred different programs . Even when voting
for the same program, some subjects found it
much better while some found it only slightly better than the other program.
Clinical Implications
The results obtained in this study showed
better speech intelligibility and listening comfort using WDRC in certain situations . This
means that there is at least some support for the
selection of single-band WDRC for clients with
moderate to moderately severe flat sensorineural
hearing loss . However, the usefulness of WDRC
in noisy situations, where it is often advertised
as being of great value, was not found .
In this study, the performance of hearing aid
programs varied with stimulus presentation
levels and SNRs . This indicates the need to
employ multiple testing levels when fitting hearing aids, especially nonlinear instruments. We
recommend that, in quiet, the hearing instruments should be evaluated with at least three
stimulus levels : high, average, and low. In noise,
SNRs of +3 and +6 dB are recommended as the
test conditions . It has been found that the average SNR of conversational speech is about +4 .8
dB in noisy environments and substantially less
in automobiles (Teder, 1990).
Subjective preference for clarity did not yield
significant findings in this study. In clinical practice, subjective evaluation of clarity may be omitted. However, the inclusion of subjective
judgments of other dimensions such as loudness
appropriateness and pleasantness may provide
some valuable information for circuit selection.
A simple paired-comparison procedure may be
adequate and easy to administer in the clinical
evaluation of hearing aid performance .
It is interesting to note that results using
materials in Cantonese, a tone language with its
own phonologic system and language structure,
yielded similar results to those obtained using
English materials . It would be interesting to
compare the results to studies using other language materials to determine whether generalizations can be made . This would have significant
implications for audiologists serving multicultural populations .
Acknowledgment . We are grateful to the Phonak
Hearing Centre Hong Kong Limited for assistance in subject recruitment and sponsorship. We thank the technical
staff in the Department of Speech and Hearing Sciences
456
for their support. Thanks also to Anthony Yuan, Polly
Lau, and Kevin Yuen for their assistance in preparing
the multitalker babble . We are indebted to the subjects
whose cooperative spirit made the research possible and
enjoyable . Appreciation is extended to Dr. Bradley
McPherson for his helpful suggestions during the preparation of this manuscript.
Kam ACS, Wong LLN. (April 4, 1998). Comparison
of Performance with Wide Dynamic Range Compression
and Linear Amplification . Poster presentation at the
10th Annual Convention of the American Academy of
Audiology [Abstract p. 1191, Los Angeles, CA .
This study was submitted by the first author for the
degree of Master of Science in Audiology at the University
of Hong Kong in May 1998.
REFERENCES
American National Standards Institute . (1989) .
Specification for Audiometers . (ANSI S3 .6-1989) . New
York : ANSI .
Bachler H, VonlanthenA . (1994) . PiCS comfort programs :
signal processing tools to support your manner of communication . Phonak Focus 17 :2-25.
Balfour PB, Hawkins DB . (1992) . Acomparison of sound
quality judgments for monaural and binaural hearing
aid processed stimuli. Ear Hear 13 :331-339 .
Boothroyd A, Springer N, Smith L, Schulman J . (1988).
Amplitude compression and profound hearing loss . J
Speech Hear Res 31 :362-376 .
Byrne D. (1996) . Hearing aid selection for the 1990s :
where to? JAm Acad Audiol 7 :377-395 .
Dillon HC . (1988) . Compression in hearing aids . In :
Sandlin RE, ed. Handbook ofHearingAidAmplification:
Vol. 1. Boston: College-Hill, 121-146 .
Dillon HC . (1996) . Compression? Yes, but for low or high
frequencies, for low or high intensities, and with what
response times? Ear Hear 17 :287-307 .
Dreschler WA. (1988a). Dynamic range reduction by peak
clipping or compression and its effects on phoneme perception in hearing-impaired listeners. Scand Audiol
17 :45-51.
Dreschler WA. (1988b) . The effect of specific compression
settings on phoneme identification in hearing-impaired
subjects . Scand Audiol 17 :35-43 .
Dreschler WA, Eberhardt D, Melk PW (1984) . The use
of single-channel compression for the improvement of
speech intelligibility. Scand Audiol 13 :231-236.
Fortune T. (1996) . Amplifiers and circuit algorithms of
contemporary hearing aids . In : Valente M, ed . Hearing
Aids : Standards, Options and Limitations . New York :
Thieme Medical, 157-209.
Fromkin V, Rodman R. (1988) . An Introduction to
Language . Fort Worth, TX : Harcourt Brace Jovanovich .
Gioannini K, Franzen R. (1978) . Comparison ofthe effects
of hearing aid harmonic distortion on performance scores
for the MRHT and PB-50 test. JAudiol Res 18 :203-208 .
Hawkins DB, Naidoo SV (1993) . Comparison of sound
quality and clarity with asymmetrical peak clipping and
Comparison of WDRC and Linear Amplification/Kam and Wong
output limiting compression . J Am Acad Audiol
4:221-237 .
pression hearing aid: paired-comparison judgments of
quality. JAcoust Soc Am 96 :1471-1478 .
Hickson LMH. (1994) . Compression amplification in hearing aids . Am J Audiol 3:51-65 .
Nilsson M, Soli SD, Sullivan JA. (1994) . Development of
the Hearing In Noise Test for the measurement of speech
reception thresholds in quiet and in noise. JAcoust Soc
Am 95 :1085-1099 .
Hickson L, Byrne D. (1995) . Acoustic analysis of speech
through a hearing aid: effects of linear vs compression
amplification . Aust J Audiol 17(1) :1-13.
Hickson L, Dodd B, Byrne D. (1995). Consonant perception with linear and compression amplification . Scand
Audiol 24 :175-184 .
Hohmann V, Kollmeier B. (1995) . The effect of multichannel dynamic compression on speech intelligibility. J
Acoust Soc Am 97 :1191-1195 .
Kuk FK. (1996) . Theoretical and practical considerations
in compression hearing aids . Trends in Amplification
1:1-39.
Ladefoged P. (1993) . A Course in Phonetics . Fort Worth,
TX : Harcourt Brace Jovanovich .
Lau CC, So KW. (1988) . Material for Cantonese speech
audiometry constructed by appropriate phonetic principles . Br J Audiol 22 :297-304 .
Lundberg G, Ovegard A, Hagerman B, Gabrielsson A,
Brandstrom U. (1992) . Perceived sound quality in ahearing aid with vented and closed earmould equalized in
frequency response . Scand Audiol 21 :87-92 .
Moore BCJ. (1996) . Perceptual consequences of cochlear
hearing loss and their implications for the design of hearing aids . Ear Hear 17 :133-160 .
Moore B, Glasberg B, Stone M. (1991) . Optimization of
a slow-acting automatic gain control system for use in
hearing aids . Br JAudiol 25 :171-182 .
Moore BCJ, Johnson JS, Clark TM, Pluvinage V (1992) .
Evaluation of a dual-channel full dynamic range compression system for people with sensorineural hearing
loss . Ear Hear 13 :349-370 .
Nabelek IV (1983) . Performance of Loaring-impaired listeners under various types of amplitude compression. J
Acoust Soc Am 74 :776-791 .
Neuman AC, Bakke MH, Hellman S, Levitt H. (1994) .
The effect of compression ratio in a slow-acting com-
Peterson M, Feeney M, Yantis P. (1990) . The effect of
automatic gain control in hearing-impaired listeners with
different dynamic ranges . Ear Hear 11 :185-194 .
Plomp R. (1988). The negative effect of amplitude compression in multi-channel hearing aids in the light of the
modulation-transfer function . J Acoust Soc Am
83 :2322-2327 .
Preves DA . (1991) . Output limiting and speech enhancement . In : Studebaker GA, Bess FH, Beck LB, eds. The
Vanderbilt Hearing Aid Report II. Parkton, MD : York
Press, 35-51.
Steinberg JC, Gardner MB . (1937) . The dependency of
hearing impairment on sound intensity. JAcoust Soc Am
9:11-23 .
Teder H. (1990) . Noise and speech levels in noisy environments . Hear Instr 41(4) :32-33 .
Tyler RS, Kuk FK. (1989) . The effects of "noise suppression" hearing aids on consonant recognition in
speech-babble and low-frequency noise. Ear Hear
10 :243-249 .
Verschuure J, Dreschler WA, de Haan EH, van Cappellen
M, Hammerschlag R, Mare MJ, Hijmans AC . (1993) .
Syllabic compression and speech intelligibility in hearing impaired listeners . Scand Audiol (Supp138):92-100.
Wong LLN, Au DKK, Tong MCF, van Hasselt A. (1996) .
Comparison of the performance of a programmable 2channel full dynamic range compression hearing aid with
1-channel conventional instruments in Hong Kong. In:
Proceedings of the Fourth Asian-Pacific Congress on
Deafness . Manila : Plantila Press, 216-224.
Wong LLN, Nilsson M, Soli SF. (in preparation) .
Deuelopment of Hearing In Noise Test-Chinese Version.
Yund E, Buckles K. (1995) . Discrimination of multi-channel-compressed speech in noise: long term learning in
hearing impaired subjects . Ear Hear 16 :417-427 .