Content uploaded by Lennart Magnusson
Author content
All content in this area was uploaded by Lennart Magnusson on Aug 05, 2014
Content may be subject to copyright.
In recent years many hearing-aid manufacturers have introduced
open-fi t hearing aids (HAs). This technology has been developed to
reduce or eliminate the occlusion effect and improve sound quality
for people with relatively normal hearing in the lower frequencies. At
present, open fi ttings constitute a considerable and increasing num-
ber of all HA fi ttings. Therefore, it is important to consider possible
limitations of this fi tting method and to be aware when a traditional
fi tting with a more closed earmold would be the better choice.
Most open-fi t HAs, like other modern HAs, are equipped with
directional microphones, which could improve speech recognition in
noise relative to performance with an omnidirectional microphone.
Previous studies have reported a signifi cantly enhanced signal-to-noise
ratio can be achieved by using directional microphones (e.g. Hawkins
& Yacullo, 1984; Valente et al, 1995; Ricketts & Dahr, 1999; Ricketts
et al, 2003). However, large earmold vents might reduce the advan-
tage of using directional microphones. Directional benefi t may be
assessed either by electroacoustical measurements or psychoacousti-
cally as the difference between speech recognition thresholds in noise
for omnidirectional and directional microphone modes. Ricketts
(2000a) evaluated the impact of venting (1 mm, 2 mm, and open) on
electroacoustically measured directivity for different hearing aids and
reported that vents decreased the directivity index (DI) of directional
HAs in the low-frequency region. Directivity at 500 Hz decreased
signifi cantly with vent size from 4.2 dB with a closed earmold to
1.9 dB with a 2 mm vent and ⫺ 2.0 dB with an open earmold. There was
also a statistically signifi cant DI reduction of about 1 dB at 1000 Hz
for all the venting conditions. The overall differences between DIs
for the omnidirectional and directional modes (i.e. directional benefi t)
were 5.6 dB with closed earmold and 4.0 dB with open earmold. This
implies that an open fi tting may impair the acoustic conditions for
directional microphones. In addition, patients intended for an open
fi tting have close to normal hearing in the lower frequencies, and
hence will receive little or no low-frequency amplifi cation, which
limits the effi cacy of any sound processing in this frequency range.
Although directional microphones might still provide front-to-back
separation in the higher frequencies, the expected improvements of
using directional microphones in open-fi t HAs can be questioned.
Valente and Mispagel (2008) examined differences between unaided
and aided performance in omnidirectional and directional modes
using open-fi t behind-the-ear HAs. Twenty-six subjects without prior
HA experience were fi t bilaterally with open-fi t HAs in which the
receiver unit were placed in the subjects ’ ear canal. The hearing
Original Article
Speech recognition in noise using bilateral open-fi t hearing
aids: The limited benefi t of directional microphones
and noise reduction
Lennart Magnusson ∗,† , Ann Claesson ∗,† , Maria Persson ∗,† & Tomas Tengstrand ∗
∗ Department of Audiology, Sahlgrenska University Hospital, Gothenburg, Sweden, and
† Department of Audiology,
Institute of Neuroscience and Physiology, University of Gothenburg, Sweden
Abstract
Objective: To investigate speech recognition performance in noise with bilateral open-fi t hearing aids and as reference also with closed earmolds, in omnidirectional mode, directional
mode, and directional mode in conjunction with noise reduction. Design: A within-subject design with repeated measures across conditions was used. Speech recognition thresholds
in noise were obtained for the different conditions. Study sample: Twenty adults without prior experience with hearing aids. All had symmetric sensorineural mild hearing loss in the
lower frequencies and moderate to severe hearing loss in the higher frequencies. Results: Speech recognition performance in noise was not signifi cantly better with an omnidirectional
microphone compared to unaided, whereas performance was signifi cantly better with a directional microphone (1.6 dB with open fi tting and 4.4 dB with closed earmold) compared
to unaided. With open fi tting, no signifi cant additional advantage was obtained by combining the directional microphone with a noise reduction algorithm, but with closed earmolds a
signifi cant additional advantage of 0.8 dB was obtained. Conclusions: The signifi cant, though limited, advantage of directional microphones and the absence of additional signifi cant
improvement by a noise reduction algorithm should be considered when fi tting open-fi t hearing aids.
Key Words: Directional benefi t; directional microphone; hearing aid; IOI-HA; Noise reduction; omnidirectional microphone;
open fi tting
Correspondence: Lennart Magnusson Department of Audiology, Sahlgrenska University Hospital, 413 45 Gothenburg, Sweden. E-mail: lennart.magnusson@neuro.gu.se
(Received 14 December 2011; accepted 20 June 2012)
ISSN 1499-2027 print/ISS N 1708-8186 online © 2013 British So ciety of Audiology, Inter nation al Societ y of Audiology, and No rdic Audiolog ical Societ y
DOI : 10.3109 /1499 2027.201 2.70 7335
International Journal of Audiology 2013; 52: 29–36
Int J Audiol Downloaded from informahealthcare.com by Sahlgrenska Universitetssjukhuset Ostra on 12/12/12
For personal use only.
30 L. Magnusson et al.
in noise test (HINT) (Nilsson et al, 1994) was used to determine
speech recognition thresholds (SRTs) in noise. The speech material
was presented from the front, and the competing noise (R-Space ™
restaurant noise) was presented via eight loudspeakers positioned
45 ° apart at a constant overall summed level of 65 dBA. Comparison
of the omnidirectional and directional results revealed a statistically
signifi cant mean directional benefi t of 1.9 dB. There was also a sta-
tistically signifi cant mean advantage of 1.7 dB for the directional
condition compared with the unaided condition. No statistically sig-
nifi cant difference was noted between the omnidirectional aided and
the unaided mean SRTs. Comparing these results with results of stud-
ies using similar loudspeaker arrangements and traditional earmold
fi tting (e.g. Pumford et al, 2000; Ricketts, 2000b) suggest that the
open-fi t approach would, to some extent, reduce the benefi t provided
by directional microphones.
A lot of HAs today can be programmed with the combination of
a directional microphone and a noise reduction (NR) algorithm. NR
refers to the ability of the HA to determine if signals are speech-like
or noise-like and make adjustments in the output of the specifi c fre-
quency region. The goal is to reduce amplifi cation in specifi c bands
when steady-state signals (e.g. background noise) are detected. How-
ever, the effi cacy of typical NR algorithms is not clear. Several stud-
ies have shown subjective benefi t of using NR algorithms but no
objective benefi t as measured with speech recognition tests (e.g.
Boymans & Dreschler, 2000; Alcantara et al, 2003; Ricketts & Hornsby,
2005). This is probably due to the fact that the competing noise often
has the same spectral shape as the speech signal. Hence, a typical NR
algorithm cannot actually improve the signal-to-noise ratio but will
reduce the gain in those frequency bands that are less important for
speech intelligibility. However, some studies have reported signifi cant
improvements of specifi c NR algorithms on objectively measured
speech recognition in noise. Peeters et al (2009) evaluated speech
intelligibility in noise offered by a commercial HA using a fully adap-
tive directional microphone and a NR algorithm that performed gain
reductions based on interactive calculation of the resulting speech
intelligibility index (SII). Eighteen subjects with varying confi gura-
tions of sensorineural hearing loss were fi t with binaural in-the-canal
HAs (N ⫽ 10) or open-fi t behind-the-ear HAs (N ⫽ 8). Speech rec-
ognition thresholds were obtained with HINT sentences in four HA
conditions; omnidirectional, omnidirectional with NR, directional,
and directional with NR. Speech material was presented directly in
front of the subjects and noise was presented from three loudspeakers
placed at 90 ° , 180 ° , and 270 ° . Results revealed an average directional
benefi t of 4 dB. Activating the NR improved the HINT results by
2.5 dB in combination with an omnidirectional microphone but only
by 0.6 dB (statistically non-signifi cant) with a directional micro-
phone. According to the authors, the limited benefi t of NR in direc-
tional mode was probably due to the fact that the signal reaching the
HA processor from the directional microphone would be lower and at
a better signal-to-noise ratio. The results were not evaluated regarding
possible differences between the open-fi t behind-the-ear HAs and
the in-the-canal HAs. Kuk et al (2005) evaluated the effi cacy of an
open-fi t HA by comparing four combinations of NR (on and off)
and microphone modes (omnidirectional and adaptive directional)
in eight subjects with open-fi t HAs. Speech recognition performance
was obtained with HINT sentences for the different conditions. Also
in this study, speech was presented directly in front of the subjects
and noise was presented from three loudspeakers placed at 90 ° , 180 ° ,
and 270 ° . The results revealed no statistically signifi cant improve-
ment with the omnidirectional microphone mode compared with
unaided when NR was not activated. However, statistically signifi cant
improvements were reported for the adaptive directional microphone
mode with and without NR. The NR algorithm improved the mean
SRT by 0.8 dB, when used in combination with an omnidirectional
or directional microphone.
Results of previous studies suggest that real-world benefi t of
directional microphones and NR algorithms depends on several addi-
tive and non-additive factors such as: type of HA, NR approach,
noise spectrum, noise source position, reverberation, and interaction
between noise management features. Particularly, the infl uence of
open fi tting on the effi cacy of directional microphones combined
with NR algorithms is not yet well established. The few reports are
diffi cult to interpret and compare due to differences in test design
(e.g. subjects, HA type, speech, and noise materials, and loudspeaker
arrangements). To determining the specifi c effect of open fi tting on
the benefi t provided by a directional microphone combined with a
NR algorithm, direct comparisons between open and closed fi ttings
are required. To the best of our knowledge, no study has previously
been published that evaluates microphone modes and NR algorithms
for an open fi tting in comparison with a traditional earmold fi tting
for the same subjects and HAs. The primary purpose of the present
study was to investigate speech recognition performance in noise
with HAs using a routinely applied open fi tting and as reference also
with a closed earmold for three different sound processing modes: (1)
omnidirectional mode, (2) directional mode, and (3) directional mode
in conjunction with NR. A secondary purpose was to evaluate the
subjective HA benefi t for a typical group of new open-fi t HA users.
The following specifi c questions were addressed:
1. Do open-fi t HAs with an omnidirectional microphone
enhance speech recognition in noise compared to unaided
performance?
2. Do directional microphones improve speech recognition in
noise in comparison with omnidirectional microphones for
open-fi t HAs?
3. Does a NR algorithm used in conjunction with a directional
microphone further improve speech recognition in noise with
open-fi t HAs?
4. To what extent might an open fi tting reduce the benefi t of
directional microphones and NR algorithms in comparison
with a closed earmold fi tting?
5. How do subjective HA outcomes of typical open-fi t HA
recipients compare to outcomes of the average HA user?
Methods
Subjects
Twenty people (10 males and 10 females), between 51 and 64 years
of age (mean: 57.3 years) without prior HA experience, participated
in this study. Considering the known variability in speech test results,
Abbreviations
ANOVA Analysis of variance
DI Directivity index
HA Hearing aid
HINT Hearing in noise test
IOI-HA International outcome inventory for hearing aids
NR Noise reduction
PB Phonemically balanced
SD Standard deviation
SRT Speech recognition threshold
Int J Audiol Downloaded from informahealthcare.com by Sahlgrenska Universitetssjukhuset Ostra on 12/12/12
For personal use only.
Performance in noise with open-fi t hearing aids 31
this sample size was estimated to be suffi cient for detecting clini-
cally signifi cant differences between conditions. The subjects were
all on the waiting list for aural rehabilitation, and were recruited
based on the following criteria: between the ages of 20 and 65 years,
Swedish as fi rst language, symmetric sensorineural hearing loss with
pure-tone thresholds of 30 dB HL or better from 250 to 1000 Hz,
and 80 dB HL or better from 2000 to 8000 Hz, speech recognition
scores with Swedish phonemically balanced (SPB) words in noise
(Magnusson, 1995) within the predicted range based on high-
frequency pure-tone thresholds and age (Barrenas & Wikstrom,
2000). Table 1 summarizes the pure-tone thresholds for both ears
of the 20 subjects. The study was approved by the local ethics
committee. The subjects provided their informed consents and
were not paid for their participation.
Hearing aid
The HA used was the Phonak Ex é lia Art M, which can be pro-
grammed as an open or closed fi tting. This HA has 20 frequency
bands that can be individually adjusted. Each subject was fi t bilat-
erally, with one pair of HAs in an open condition with thin tubes
and open domes, and with another pair of HAs in a closed condi-
tion using unvented shell molds made of acrylic material. The HAs
were programmed based on the audiometric confi guration of each
subject and the fi tting approach (open or closed) using a “ fi rst fi t ”
with the Phonak adaptive digital formula. Phonak recommends using
this prescription formula “ rather than distinct separate fi tting formu-
las ” . The fi tting formula is protected by a patent, but a description
provided by the manufacturer is presented in the Appendix. The
fi ttings were performed in accordance with local clinical routine for
fi tting this type of HA. That is, the manufacturer ’ s fi tting formula
was used and no real ear measurements were performed. Separate
programs were made for the three processing modes: (1) omnidi-
rectional microphone mode, (2) directional microphone mode, and
(3) directional microphone mode in conjunction with NR. A fi xed
supercardioid polar pattern was chosen in the directional modes.
The level of NR (weak, moderate, or strong) was set to moderate.
According to information from Phonak, their NR algorithm is based
on the signal-to-noise ratio in the specifi c channel where the noise
appears, and the moderate setting provides an improved speech-to-
noise ratio for speech in speech-shaped unmodulated noise of 2 dB at
a 0 dB signal-to-noise ratio and 4 dB at a ⫺ 5 dB signal-to-noise ratio.
The HAs were not fi ne-tuned until data collection was fi nished.
Speech recognition test
The Swedish speech test material, Hagerman ’ s sentences (Hagerman,
1982), was used to determine SRTs in noise. This test was originally
developed in Swedish but is now available in several other languages.
Each of the eleven lists in this test material consists of ten sentences
with fi ve words. All sentences have the same structure: name –
verb – numeral – adjective – noun (e.g. “ Peter has four black bas-
kets ” ). Hagerman (1982) developed this test material by recording
one original list with a female speaker, and then he compiled the var-
ious lists by combining the words differently between the sentences.
Because all lists comprise exactly the same 50 recorded words,
high test-retest reliability is achieved. According to Hagerman and
Kinnefors (1995) the standard deviation of repeated measurements
is 0.44 dB including the learning effect. Another advantage of these
low-predictability sentences is that the lists can be used repeatedly
with the same subject, because it is almost impossible to learn the
lists by heart (Wagener et al, 2003). Hagerman ’ s sentence material
is used in conjunction with a specifi c masker (i.e. noise signal) that
has the same long-term spectrum as the speech signal. The masker is
available in a slightly (10%) and a fully (100%) amplitude modulated
version. In the present study, the slightly modulated version was used
in order to activate the NR algorithm in the HAs.
Speech tests were performed in an audiometric test room at Sahl-
grenska University Hospital. The speech and noise materials were
played from a CD-player (Philips CD620) connected to an audiom-
eter (Interacoustics AC30) that was calibrated according to ISO-389.
The subject was seated in the middle of the room facing a loud-
speaker that presented the speech signal. The noise signal was routed
through an Interacoustics Directional Hearing Evaluator DHA8 and
presented uncorrelated via four other loudspeakers positioned at 45 ° ,
135 ° , 225 ° , and 315 ° around the subject. All the loudspeakers were
of the same model, Bose interaudio 1000XL, and were positioned at
the height of the subject ’ s head at a distance of 1 m. Calibration of the
signals was performed in the test position (i.e. a point corresponding
to the center of the subject ’ s head).
The sentences were presented at a fi xed level of 65 dB SPL,
and the level of the competing noise was adapted according to the
method described by Hagerman and Kinnefors (1995). The noise
level was initially set at 45 dB SPL and was increased in 5-dB steps
until two words or less was repeated correctly. Thereafter, the noise
was adjusted in 1-, 2- or 3-dB steps according to the adaptive step-
ping scheme outlined in Table 2. This stepping scheme will give
an SRT that converges at a speech-to-noise ratio corresponding to
40% correct. One list takes less than two minutes to complete.
Follow-up questionnaire
Four to six weeks after completing the experimental part of the study,
the benefi t of the subjects ’ open-fi t HAs was subjectively determined
with the international outcome inventory for hearing aids (IOI-HA).
This questionnaire, which originally was developed in English (Cox
et al, 2000), has been translated to several other lang uages, including
Swedish (Cox et al, 2002). Psychometric properties of the Swedish
version have recently been determined (Brannstrom & Wennerstrom,
2010). IOI-HA comprises seven questions, each assessing a different
self-report outcome dimension. These dimensions are: (1) daily use,
(2) increased activity, (3) residual activity limitation, (4) satisfaction,
(5) residual participation restriction, (6) impact on others, and (7)
quality of life. Each question has fi ve different response choices,
from 1 (worst) to 5 (best), which always proceed from the worst
Table 1. Mean, standard deviation (SD), minimum, and maximum
values of the hearing threshold levels for the 20 subjects; results for
the right and left ears are combined.
Frequency (Hz) 250 500 1000 2000 3000 4000 6000 8000
Mean (dB HL) 14.5 16.3 20.1 35.6 42.5 48.6 53.2 57.1
SD (dB) 7.8 9.0 8.4 10.1 8.8 8.5 12.3 14.2
Minimum (dB HL) 0001025302030
Maximum (dB HL) 30 30 30 60 60 70 80 80
Table 2. The adaptive stepping scheme used for Hagerman ’ s speech
test (Hagerman & Kinnefors, 1995) .
Number of correctly repeated words 012345
Change of noise level (dB) ⫺ 2 ⫺ 10 ⫹ 1 ⫹ 2 ⫹ 3
Int J Audiol Downloaded from informahealthcare.com by Sahlgrenska Universitetssjukhuset Ostra on 12/12/12
For personal use only.
32 L. Magnusson et al.
outcome on the left side to the best outcome on the right side of
the form.
Procedure
The present study comprised three appointments spaced 4 to 6 weeks
apart. At the fi rst session ear impressions were made for the ear-
molds, and subsequently the unaided SRT in noise was obtained
with Hagerman ’ s sentences using two scoring lists (20 sentences)
preceded by a practice list to fi nd the starting level for the noise.
The subjects were informed that the sentences would consist of fi ve
words and that the noise level would vary, which makes it sometimes
hard and sometimes easy to perceive the sentences. The subjects
were instructed to guess when they could not clearly perceive the
spoken words.
At the second session, four to six weeks later, two binaural HA
fi ttings were carried out, one open fi tting and one with closed ear-
molds. After each fi tting, the SRTs in noise were obtained for the
three processing modes: (1) omnidirectional microphone mode,
(2) directional microphone mode, and (3) directional microphone
mode in conjunction with NR. In order to avoid learning effects,
every other subject started with open fi tting and the others started
with closed earmolds, and the sequence of the three processing modes
was counterbalanced across the subjects according to a cyclic permu-
tation. After an initial practice list, SRTs were determined using two
lists (20 sentences) for each condition and microphone mode. Thus,
each subject had to listen to 13 lists at the second session. Pauses
were taken when changing between the open-fi t HAs and those with
earmold. After completing the speech recognition tests, the subjects
went home with the open-fi t HAs and four selectable programs: (1)
“ Soundfl ow ” , (2) omnidirectional mode, (3) directional mode, and
(4) directional mode in conjunction with NR. “ Soundfl ow ” is the
standard adaptive program in the Phonak Exelia M, which adapts
automatically to environments such as quiet, speech in noise, com-
fort in noise, and music.
At the third session, another four to six weeks later, subjects eval-
uated if they wanted to continue to use these HAs, fi ne tune the HAs,
test other HAs or discontinue using HAs. The IOI-HA questionnaire
was also completed at the third session to assess subjective benefi t
of the open-fi t HAs.
Results
Speech recognition thresholds
Figure 1 shows the means and standard deviations (SD) of the
SRTs obtained for the different conditions and microphone modes.
Because the SRTs are expressed as speech-to-noise ratios, a lower
(more negative) SRT indicates better performance. The mean
unaided SRT was ⫺ 6.0 dB (SD ⫽ 2.0 dB). With open-fi t HAs, the
mean SRTs were: ⫺ 6.3 dB (SD ⫽ 1.5 dB) in omnidirectional mode,
⫺ 7.6 dB (SD ⫽ 1.3 dB) in directional mode, and ⫺ 7.9 dB (SD ⫽ 1.3
dB) in directional mode in conjunction with NR. The mean SRTs
with closed earmolds were: ⫺ 7.1 dB (SD ⫽ 1.2 dB) in omnidirec-
tional mode, ⫺ 10.4 (SD ⫽ 1.4 dB) in directional mode, and ⫺ 11.2
(SD ⫽ 1.3 dB) in directional mode in conjunction with NR.
A two-way repeated measures ANOVA conducted on the SRTs
for the aided conditions reveled signifi cant main effects of fi tting
(open, closed) [F(1, 19) ⫽ 73.9, p ⬍ 0.001] and microphone mode
(omnidirectional, directional, directional with NR) [F(1, 19) ⫽ 54.9,
p ⬍ 0.001]. The interaction between fi tting and microphone mode
was also signifi cant [F(1, 19) ⫽ 29.0, p ⬍ 0.001].
Differences between the mean SRTs for all listening conditions
and microphone modes are presented in Table 3. Post-hoc Bonfer-
roni adjusted pairwise comparisons revealed statistically signifi -
cant differences between most conditions as indicated in Table 3.
However, there were no signifi cant differences between the unaided
SRTs and the SRTs obtained using HAs with an omnidirectional
microphone, regardless if the fi ttings were open or closed. Using
open-fi t HAs with directional microphones yielded a 1.6 dB sig-
nifi cantly better SRT than unaided (p ⫽ 0.01). The directional mode
yielded signifi cantly better SRT than the omnidirectional mode with
the open and closed fi tting (p-values ⬍ 0.001). Activating the NR in
directional mode did not provide a signifi cant additional improve-
ment of the SRT when using an open fi tting. However, with closed
earmolds a 0.8-dB signifi cant additional improvement (p ⫽ 0.004)
was achieved.
Subjective follow-up
The subjective outcome of the open HA fi tting for the study subjects
was evaluated after one month using the IOI-HA questionnaire. The
results presented in Table 4 are based on 18 subjects. One subject
discontinued HA use due to problems in the ear canal, and another
subject did not complete the questionnaire due to lack of motivation.
The total mean score was 28.9 (SD ⫽ 4.7). The highest score was
obtained on item 6 (i.e. impact on others). For comparison, Table 4
also provides results from a recent psychometric evaluation of the
Swedish version of IOI-HA (Brannstrom & Wennerstrom, 2010). On
item 1, which refl ects daily use, nine subjects reported a usage of
more than 8 hours/day, and six subjects reported a usage between 4
and 8 hours/day. None of the 18 subject reported a HA use of less
than 1 hour/day. Seventeen of the 20 subjects continued with the
actual open-fi t HAs after the third visit, one decided to test HAs from
another manufacturer, and two decided not to continue with HAs.
Discussion
The primary purpose of this study was to examine speech recogni-
tion performance in noise with open-fi t HAs in the omnidirectional
mode, directional mode, and a combined directional and NR mode.
Figure 1. Mean speech recognition threshold (SRT (dB)) in noise
for the different listening conditions: unaided, open omnidirectional
(Omni), open directional (Dir), open directional with noise reduction
(Dir/NR), closed omnidirectional, closed directional and closed
directional with noise reduction. Error bars indicate ⫾ 1 SD.
Int J Audiol Downloaded from informahealthcare.com by Sahlgrenska Universitetssjukhuset Ostra on 12/12/12
For personal use only.
Performance in noise with open-fi t hearing aids 33
To facilitate interpretation of the results and enable further com-
parisons between conditions, reference SRTs were also obtained
with closed earmolds and unaided. The results revealed a signifi -
cant improvement of speech recognition in noise when using direc-
tional microphones compared with omnidirectional. However, the
directional advantage was on average 1.9 dB smaller for the open
fi ttings compared with closed. Activating the NR feature yielded a
signifi cant additional improvement, but only with closed earmolds.
Compared with unaided, there was no signifi cant advantage of using
HAs with omnidirectional microphones in background noise, neither
with open nor with closed fi tting.
The test subjects had no prior experience with HAs, and SRTs
were obtained for all conditions without any time for adapting to
the HAs. An acute testing procedure was considered appropriate,
because more acclimatization to some of the test conditions would
have put those conditions in favor. Thus, all tested conditions,
except the unaided, represented unfamiliar listening situations for
all subjects. The study was conducted using one specifi c HA model
(Phonak Ex é lia Art M) and the fi ttings were performed using the
Phonak adaptive digital formula. This HA is suitable both for open
fi tting and for fi tting with a traditional earmold. To further minimize
variations of different parameters in the study, no fi ne-tuning of
the HAs were made before the data collection was completed. It
is important to consider that one HA model was used and fi ttings
were performed in accordance with local clinical routine using the
actual manufacturer ’ s fi tting formula without any real-ear measure-
ments. Thus, the fi ttings may not have been optimal for all subjects
and conditions. However, according to a comprehensive fi eld study
reported by Phonak, fi nal settings preferred by users did not differ by
more than 3 – 4 dB in single channels and on average only by 1 – 2 dB
from the settings prescribed by the Phonak adaptive digital formula
(Hessefort, 2010). When fi tting the actual HA, like many other HAs,
an adaptive directional microphone mode is commonly selected; this
means that the microphone adapts its directionality depending on
where the noise is coming from. However, in this study noise was
presented with equal levels from four loudspeakers around the sub-
ject. Therefore, a fi xed frontal supercardioid polar pattern was used
to avoid random changes of polar pattern throughout the data collec-
tion, which in turn could have affected the reliability of the results.
When generalizing laboratory test results to real-world situations,
valid test materials and methods are needed. However, for the cur-
rent purpose of making relative comparisons between conditions,
high test-retest reliability was considered more important than maxi-
mal ecologic validity. Reliability of speech recognition tests can
be expressed as the SD for repeated measurements. The reported
normative (i.e. for normal-hearing subjects) SDs for repeated mea-
surements are 0.44 dB for Hagerman ’ s sentences (Hagerman &
Kinnefors, 1995), and 0.96 dB for the Swedish HINT (Hallgren et al,
2006). Therefore, we used Hagerman ’ s sentences, although these
sentences are somewhat unnatural in comparison with everyday
sentences such as the HINT sentences.
The present results revealed SRTs that ranged between ⫺ 6.0
and ⫺ 11.2 dB for the different conditions and fi tting methods. The
relationship between the SRTs for the unaided, the omnidirectional,
and directional open-fi t conditions were in good agreement with
those obtained in the Valente and Mispagel (2008) study. The aver-
age directional advantage with open fi ttings in the current study was
1.4 dB, which is slightly lower than the advantage of 1.9 dB reported
by Valente and Mispagel (2008) who used a more diffi cult listen-
ing environment with noise from eight loudspeakers from the front,
sides, and behind. In the current study noise was presented from
four loudspeakers equally spaced around the subject, but not directly
from the front. One might expect greater advantage of directional
microphones when less noise is presented from the front; however
this situation is also favorable to the omnidirectional microphone
mode because of the negative directionality of behind-the-ear HAs
in omnidirectional mode (e.g. Ricketts, 2000a). There might also
have been differences in openness, because Valente and Mispagel
(2008) used HAs with the receiver placed in the ear canal, while thin
tubes and domes were used in the current study. There are very few
other publications reporting on the magnitude of directional benefi t
of open-fi t HAs. Flynn (2004) presented a study with 49 experienced
HA users, of whom 38 were using in-the-ear HAs with an average
vent size of 2.3 mm, and 11 were using behind-the-ear HAs with an
Table 3. Differences between the SRTs (dB) for combinations of the unaided, open and closed conditions,
and the omnidirectional (Omni), directional (Dir), and directional ⫹ noise reduction (Dir ⫹ NR) modes.
The levels of statistical signifi cance for the differences are marked with stars.
Condition Unaided Open/Omni Open/Dir Open/Dir ⫹ NR Closed/Omni Closed/Dir
Open/Omni 0.22
Open/Dir 1.57
∗ ∗ 1.35
∗ ∗ ∗
Open/Dir ⫹ NR 1.84
∗ ∗ 1.63
∗ ∗ ∗ 0.27
Closed/Omni 1.10 0.89 0.47 0.74
Closed/Dir 4.38
∗ ∗ ∗ 4.17
∗ ∗ ∗ 2.81
∗ ∗ ∗ 2.54
∗ ∗ ∗ 3.28
∗ ∗ ∗
Closed/Dir ⫹ NR 5.14
∗ ∗ ∗ 4.92
∗ ∗ ∗ 3.57
∗ ∗ ∗ 3.30
∗ ∗ ∗ 4.04
∗ ∗ ∗ 0.76
∗ ∗
∗ p ⬍ 0.05; ∗ ∗ p ⬍ 0.01; ∗ ∗ ∗ p ⬍ 0.001.
Table 4. Mean results (with standard deviation within parentheses) of the IOI-HA questionnaire. The
items refl ect the following domains: (1) daily use, (2) increased activity, (3) residual activity limitation,
(4) satisfaction, (5) residual participation restriction, (6) impact on signifi cant others, and (7) quality of
life. Reference values are for a Swedish validation of IOI-HA (Brannstrom & Wennerstrom, 2010) .
Item number 1234567
Present study 4.2 (0.8) 4.1 (1.2) 3.8 (0.9) 4.2 (1.1) 4.3 (0.9) 4.6 (0.6) 3.8 (1.0)
Reference 3.9 (1.1) 4.0 (1.1) 3.5 (1.2) 4.3 (1.0) 4.1 (1.1) 3.9 (1.1) 3.8 (1.0)
Int J Audiol Downloaded from informahealthcare.com by Sahlgrenska Universitetssjukhuset Ostra on 12/12/12
For personal use only.
34 L. Magnusson et al.
average vent size of 3.9 mm. SRTs were obtained with the Danish
version of Hagerman ’ s sentences denoted DANTALE II (Wagener
et al, 2003). The speech signal was presented from the front, while
noise was presented via four loudspeakers located behind the subject
at 120 ° , 150 ° , 210 ° , and 240 ° . Flynn reported a signifi cant mean
directional benefi t of 3.1 dB. Kuk et al (2005) presented results from
eight subjects using open-fi t behind-the-ear HAs. The subjects were
tested with HINT sentences presented from the front, and noise pre-
sented at 90 ° , 180 ° and 270 ° . The results indicated a directional
benefi t of 1.8 dB. Thus, the directional benefi t with open-fi t HAs
seems not to exceed 2 dB in situations where noise is presented from
all directions. The markedly higher benefi t reported by Flynn (2004),
was probably due to the fact that the fi ttings were not completely
open and the noise signal was only presented from behind.
Regardless of open or closed fi tting, the effect of directional
microphones might differ between HA manufacturers, HA models,
and even the microphone position, because the DI can decrease if
the alignment angle varies by more than 10 to 15 ° in the horizontal
plane (Ricketts et al, 2003). In addition, results of laboratory tests are
infl uenced by differences in test procedures. The results of previous
studies, in which the researchers have used similar speech test mate-
rials and loudspeaker arrangements for estimating directional benefi t
with traditionally fi tted HAs, range from 2.0 to 4.9 dB (e.g. Ricketts,
2000a; Ricketts & Mueller, 2000; Bentler et al, 2004). Thus, open
fi ttings probably reduce directional advantage to some extent com-
pared with earmold fi ttings. The current study was design to enable
direct comparison between an open fi tting and an earmold fi tting for
the same subjects and HAs. The mean directional advantage with
closed earmolds was 3.3 dB, which is within the range of the previ-
ous results mentioned above and 1.9 dB higher than the directional
advantage we obtained with an open fi tting. Thus, using open fi tting
can reduce directional benefi t of HAs to 40% in comparison with
earmold fi tting. This knowledge would be helpful especially when
counseling people who are close to the border of the open fi tting
range. The maximum performance of the directional microphones in
the present HAs was established with completely closed earmolds.
Of course completely closed earmolds would not be a real option
for the actual population, but small vents do not materially affect
directional advantage (Ricketts & Mueller, 2000). For people need-
ing a more closed fi tting because of a signifi cant amount of hearing
loss in the low-frequency region, a partially occluding earmold can
be connected to the thin tube. In recent years hollow earmolds have
become popular in combination with thin-tube fi tting. An advantage
of hollow earmolds is that that similar vent effects can be achieved
with a smaller vent diameter in comparison with a solid earmold
(Kuk et al, 2009). However, studies are needed to evaluate possible
differences between hollow and solid earmolds regarding the effect
of directional microphones and NR algorithms.
The other main objective of this study was to assess the ben-
efi t provided by the HA ’ s NR algorithm when used in combina-
tion with the directional microphone. It is important to validate
this combination in open-fi t HAs, given that most manufacturers
recommend using a directional microphone and NR in the same
program. As mentioned in the introduction, Kuk et al (2005)
reported that a NR algorithm improved the mean SRT by 0.8 dB,
when used in combination with an omnidirectional or directional
microphone in open-fi t HAs. The present results showed a 0.3 dB
non-signifi cant additional improvement of NR when using an open
fi tting, but with closed earmolds a statistically signifi cant advantage
of 0.8 dB was obtained. There might be considerable differences in
NR between HA brands and models, due to differences in sound
processing and implementation of NR algorithms. For the HAs in
this study, we used the middle (denoted moderate) of three selectable
levels of the NR, and possibly, the highest level would have gener-
ated greater improvement. However, according to data provided by
Phonak, the moderate setting should improve SRTs by about 4 dB
using unmodulated speech-shaped noise at speech-to-noise ratios
around ⫺ 5 dB. Hagerman ’ s noise, which was used in the current
study, is speech-shaped but also slightly (10%) modulated. Possibly,
this slight modulation would have caused the NR algorithm not to
activate. However, it did activate because there was a signifi cant
effect of NR with closed earmolds. It has been shown that a NR
algorithm might be less effective when used in combination with
a directional microphone. Peeters et al. (2009) reported signifi cant
improvement for a NR algorithm when used in combination with an
omnidirectional microphone, but little or no improvement in com-
bination with a directional microphone. In the present study the NR
was only evaluated in combination with a directional microphone
because the manufacturer recommends using this combination. The
most probable reason for the small effect of NR in the current study
is that the actual speech test material, Hagerman ’ s sentences, has
identical frequency spectra for speech and noise, hence there were
no specifi c frequency bands in which more noise could be detected
and gain be reduced. Therefore, the low-frequency gain was prob-
ably reduced, which provided a small benefi t with closed earmolds
but not with open fi tting where the noise was transmitted directly
into the open ear canal. It should be pointed out that a test material
with similar speech and noise spectra and a slightly modulated noise
signal indeed is valid for the daily challenge of speech recognition
in multi-talker surroundings.
The present results, in accordance with previous results (Kuk
et al, 2005; Valente & Mispagel, 2008), have shown that people
with normal hearing or a mild hearing loss in the low-frequency
region and a moderate hearing loss in the high-frequency region
will receive benefi t from using open-fi t HAs with directional micro-
phones in noisy situations. On the contrary, open-fi t HAs with omni-
directional microphones do not improve performance compared with
the unaided performance in a noisy environment at normal listening
levels. Further, the current results suggest that using NR algorithms
in combination with directional microphones might not provide
additional benefi t for open-fi t HA users in many common listening
situations.
The actual patient population generally has minor diffi culty in com-
municating with others in quiet environments, whereas their hearing
problems often become apparent in a noisy situation. Therefore, it
is important to give priority to directional mode in open-fi t HAs,
even though the directional advantage would be smaller than for
earmold fi ttings. Omnidirectional mode could be preferable in some
situations, for example, when listening to soft sounds from different
directions. However, the effi cacy of NR algorithms in combination
with omnidirectional microphones needs to be evaluated for open fi t-
tings. The present results indicated limited but signifi cant advantages
of using open-fi t HAs compared with the unaided condition for the
actual population, however, users might benefi t even more in real
life. It is important to consider the limitations of laboratory settings
where only one type of speech material, one specifi c noise signal,
one loudspeaker arrangement, and one listening level are being used.
It should also be pointed out, that presenting a difference between
conditions as a speech-to-noise ratio in dB, which is a more abstract
measure than a percentage score, might lead to underrating of the
real-world impact. An improved speech-to-noise ratio of 1 dB cor-
responds to an improved recognition score of up to 25 percentage
Int J Audiol Downloaded from informahealthcare.com by Sahlgrenska Universitetssjukhuset Ostra on 12/12/12
For personal use only.
Performance in noise with open-fi t hearing aids 35
points (pp) for Hagerman ’ s sentence test, which has a very steep
psychometric function (Hagerman, 1982). Although the slopes of the
psychometric functions for speech materials that are comprised by
everyday sentences are commonly shallower, e.g. American English
HINT (Nilsson et al, 1994): 9.7 pp/dB, and Swedish HINT (Hallgren
et al, 2006): 17.9 pp/dB, it is clear that just one dB improvement
of the speech-to-noise ratio would make great difference in speech
recognition performance.
After fi nishing the experimental part of the study the subjects
went home with their open-fi t HAs. At a follow-up appointment
four to six weeks later their HA outcomes were subjectively evalu-
ated using the Swedish version of IOI-HA. A psychometric evalu-
ation of the Swedish version of this questionnaire has recently been
performed (Brannstrom & Wennerstrom, 2010). This evaluation
comprised 224 people with different types of hearing losses and
no prior HA experience, who received either unilateral or bilateral
HAs. Different kinds of HAs were used: behind the ear, completely
in the canal, and open-fi t HAs. Six months after the fi tting was com-
pleted, the subjects evaluated their HAs by completing the IOI-HA
questionnaire. The results were in accordance with results obtained
with other versions of IOI-HA. The global mean score was 27.7
(SD ⫽ 5.2), and the results for each item are presented as refer-
ence values in Table 4. The global mean score for the current test
subjects after 4 – 6 weeks with open-fi t HAs was 28.9 (SD ⫽ 4.7).
The scores on the different items were equal or slightly higher
than the scores reported in the evaluation study (Brannstrom &
Wennerstrom, 2010). The greatest difference was found on item
6 (impact on others), where the mean score was 4.6 as compared
with 3.9. Also, item 1 (usage) yielded a higher mean score in the
present study, 4.2 as compared with 3.9. The short period of time
before administering the questionnaire (i.e. 1 month vs. 6 months)
could have infl uenced the results. However, IOI-HA scores have
shown to be quite stabile over time (Vestergaard, 2006). Thus the
results indicate that the present subjects, despite only mild to mod-
erate hearing loss, were on average at least as satisfi ed with their
HAs as are most HA users.
Conclusions
The present results confi rm that open fi tting signifi cantly reduces
the possible benefi t of directional microphones and NR algorithms
in HAs. More studies are needed to further evaluate the effi cacy
of open-fi t HAs and compare different designs, technologies and
special algorithms. It is also important to expand our knowledge
about the most appropriate fi tting method for different audiometric
confi gurations, and establish guidelines for counseling patients to
the best choice. Because the unaided performance in noise may not
be improved by using open-fi t HAs with omnidirectional micro-
phones, it is important to give priority to directional mode in open-
fi t HAs.
Based on the current results, while also considering previous
studies, we make the following specifi c conclusions regarding HAs
for people with normal hearing or a mild hearing loss in the low-
frequency region and a moderate to severe hearing loss in the high-
frequency region:
1. Using open-fi t HAs with omnidirectional microphones do not
signifi cantly enhance speech recognition in noise compared to
unaided performance.
2. Using open-fi t HAs with directional microphones can sig-
nifi cantly improve speech recognition in noise compared to
unaided and omnidirectional aided performance.
3. Utilizing NR algorithms in combination with directional micro-
phones in open-fi t HAs may not further improve speech rec-
ognition in noise.
4. Advantages of directional microphones and NR algorithms are
signifi cantly reduced by using open fi tting compared to con-
ventional earmolds
5. The results of the IOI-HA questionnaire indicate that successful
hearing rehabilitation of people with this common type of hear-
ing loss can be achieved by means of open-fi t HAs.
Acknowledgements
The authors would like to thank Michael Valente, Kristi Oeding, and
an anonymous reviewer for their important comments and helpful
suggestions on an earlier version of the manuscript.
Declaration of interest: The authors report no confl ict of interest.
The authors alone are responsible for the content and the writing of
the paper.
The study was supported by a grant from the Rune and Ulla
Aml ö vs foundation for audiological research.
References
Alcantara J.L. , Moore B.C. , Kuhnel V. & Launer S . 2003 . Evaluation of the
noise reduction system in a commercial digital hearing aid . Int J Audiol ,
42 , 34 – 42 .
Barrenas M.L. & Wikstrom I . 2000 . The infl uence of hearing and age on
speech recognition scores in noise in audiological patients and in the
general population . Ear Hear , 21 , 569 – 577 .
Bentler R.A. , Egge J.L. , Tubbs J.L. , Dittberner A.B. & Flamme G.A . 2004 .
Quantifi cation of directional benefi t across different polar response
patterns . J Am Acad Audiol , 15 , 649 – 659 .
Boymans M. & Dreschler W.A . 2000 . Field trials using a digital hearing
aid with active noise reduction and dual-microphone directionality .
Audiology , 39 , 260 ⫺ 268 .
Brannstrom K.J. & Wennerstrom I . 2010 . Hearing-aid fi tting outcome: Clini-
cal application and psychometric properties of a Swedish translation
of the international outcome inventory for hearing aids (IOI-HA) . J Am
Acad Audiol , 21 , 512 – 521 .
Cox R. , Hyde M. , Gatehouse S. , Noble W. , Dillon H. et al . 2000 . Optimal
outcome measures, research priorities, and international cooperation .
Ear Hear , 21 , 106S ⫺ 115S .
Cox R.M. , Stephens D. & Kramer S.E . 2002 . Translations of the International
Outcome inventory for Hearing Aids (IOI-HA) . Int J Audiol , 41 , 3 – 26 .
Flynn M . 2004 . Maitaining the directional advantage in open fi ttings . Hear
Rev , 11 , 32 – 36 .
Hagerman B . 1982 . Sentences for testing speech intelligibility in noise . Scand
Audiol , 11 , 79 – 87 .
Hagerman B. & Kinnefors C . 1995 . Effi cient adaptive methods for measur-
ing speech reception threshold in quiet and in noise . Scand Audiol , 24 ,
71 – 77 .
Hallgren M. , Larsby B. & Arlinger S . 2006 . A Swedish version of the hear-
ing in noise test (HINT) for measurement of speech recognition . Int J
Audiol , 45 , 227 – 237 .
Hawkins D.B. & Yacullo W.S . 1984 . Signal-to-noise ratio advantage of
binaural hearing aids and directional microphones under different levels
of reverberation . J Speech Hear Disord , 49 , 278 – 286 .
Hessefort K . 2010 . High spontaneous acceptance and excellent customer
satisfaction. Field Study News : Phonak hearing systems .
Kuk F. , Keenan D. & Lau C.C . 2009 . Comparison of vent effects between a
solid earmold and a hollow earmold . J Am Acad Audiol , 20 , 480 – 491 .
Kuk F. , Keenan D. & Ludvigsen C . 2005 . Effi cacy of an open-fi tting hearing
aid . Hear Rev , 12 , 26 – 31 .
Int J Audiol Downloaded from informahealthcare.com by Sahlgrenska Universitetssjukhuset Ostra on 12/12/12
For personal use only.
36 L. Magnusson et al.
Magnusson L . 1995 . Reliable clinical determination of speech recognition
scores using Swedish PB words in speech-weighted noise . Scand Audiol ,
24 , 217 – 223 .
Nilsson M. , Soli S.D. & Sullivan J.A . 1994 . Development of the hearing in
noise test for the measurement of speech reception thresholds in quiet
and in noise . J Acoust Soc Am , 95 , 1085 – 1099 .
Peeters H. , Kuk F. , Lau C.C. & Keenan D . 2009 . Subjective and objective eval-
uation of noise management algorithms . J Am Acad Audiol , 20 , 89 – 98 .
Pumford J.M. , Seewald R.C. , Scollie S.D. & Jenstad L.M . 2000 . Speech
recognition with in-the-ear and behind-the-ear dual-microphone hearing
instruments . J Am Acad Audiol , 11 , 23 – 35 .
Ricketts T . 2000a . Directivity quantifi cation in hearing aids: Fitting and
measurement effects . Ear Hear , 21 , 45 – 58 .
Ricketts T . 2000b . Impact of noise source confi guration on directional
hearing aid benefi t and performance . Ear Hear , 21 , 194 – 205 .
Ricketts T. & Dahr S . 1999 . Comparison of performance across three
directional hearing aids . J Am Acad Audiol , 10 , 180 – 189 .
Ricketts T. , Henry P. & Gnewikow D . 2003 . Full time directional versus user
selectable microphone modes in hearing aids . Ear Hear , 24 , 424 – 439 .
Ricketts T. & Mueller H.G . 2000 . Predicting directional hearing-aid benefi t
for individual listeners . J Am Acad Audiol , 11 , 561 – 569 .
Ricketts T.A. & Hornsby B.W . 2005 . Sound quality measures for speech
in noise through a commercial hearing aid implementing digital noise
reduction . J Am Acad Audiol , 16 , 270 – 277 .
Wagener K. , Josvassen J.L. & Ardenkjaer R . 2003 . Design, optimization, and
evaluation of a Danish sentence test in noise . Int J Audiol , 42 , 10 – 17 .
Valente M. , Fabry D.A. & Potts L.G . 1995 . Recognition of speech in
noise with hearing aids using dual microphones . J Am Acad Audiol , 6 ,
440 – 449 .
Valente M. & Mispagel K.M . 2008 . Unaided and aided performance with a
directional open-fi t hearing aid . Int J Audiol , 47 , 329 – 336 .
Vestergaard M.D . 2006 . Self-report outcome in new hearing-aid users:
Longitudinal trends and relationships between subjective measures of
benefi t and satisfaction . Int J Audiol , 45 , 382 – 392 .
Appendix
Calculation of targets: Phonak Adaptive Digital
Rather than distinct, separate fi tting formulas, the hearing instrument
uses the Phonak Adaptive Digital fi tting formula, which includes
transitions for different hearing loss confi gurations. As such it recog-
nizes confi gurations such as ‘ standard ’ , ‘ ski-slope ’ ‘ low-frequency ’
and ‘ profound ’ hearing losses. The Phonak Adaptive Digital fi tting
formula also recognizes the importance of reaching gain targets at
specifi c frequencies, depending on the hearing loss confi guration.
For example, the importance of reaching targets for speech-relevant
inputs and frequencies (the so-called speech banana) is weighted
higher than for frequencies outside this speech banana. The transi-
tions from ‘ standard ’ to a different fi tting pre-calculation are:
T RANSITION TO SKI-SLOPE CRITERIA
• Hearing thresholds at or below 750 Hz to be less than 30 dB HL.
• Hearing thresholds at 4000 Hz or higher should be at least
50 dB more than hearing thresholds at 500 Hz or lower.
• The slope over two octaves between 500 Hz and 4000 Hz inclu-
sive should be at least 50 dB.
T RANSITION TO LOW-FREQUENCY HEARING LOSS CRITERIA
• Hearing thresholds at or below 750 Hz should be at least 25 dB
worse than at 1500 Hz or higher.
• Hearing thresholds between 2000 and 4000 Hz (inclusive)
should be 20 dB HL or better.
• Weight shapes are low for regions with hearing loss of more
than 45 dB HL over the minimum hearing loss between 500
and 4000 Hz.
T RANSITION TO PROFOUND HEARING LOSS CRITERIA
• Average hearing thresholds at 500, 1000, 2000, and 4000 Hz
should be at least 75 dB HL.
• Hearing thresholds should be more than 55 dB HL.
Int J Audiol Downloaded from informahealthcare.com by Sahlgrenska Universitetssjukhuset Ostra on 12/12/12
For personal use only.
