Transcript Improved acoustic parameters

21st Danavox Symposium
31 Aug-2 Sept 2005
Kolding, Denmark
The effect of advanced
signal processing strategies
in hearing aids on user
performance and preference
Gitte Keidser, Lyndal Carter, and
Harvey Dillon
National Acoustic Laboratories
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Introduction

Modern hearing aids contain a DSP
computer and are software programmable
 complex
and multiple manipulations of sound
 precise and flexible adjustments
 automated adjustments to specific client data
To what extent do the advances in hearing
aid technology benefit the hearing aid
user?
 Focus on clinical implications

National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Sound quality comparisons



Digital signal processing introduces new forms
of distortion in hearing aids, e.g. due to analysis
of sound into different frequency regions and
subsequent resynthesis
A new processing strategy operating in the timedomain using channel-free signal processing
has been introduced
Does the sound quality differ in advanced
hearing aids and can objective measurements
predict the subjective preference?
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Sound quality comparisons

Test devices
 Canta
7 (770-D)
 Claro 211 dAZ
 Senso Diva SD-9M
 Symbio 100
 Triano S
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Objective measures
Poorer than average
Coherence value
Time delay (msec)
Harmonic distortion (%)
Internal noise (dB)
Peak MPO (dB)
Standardised normalised
values
4
3
p = 0.50
2
1
0
-1
-2
Canta
Better than average
Claro
Diva
Symbio
Triano
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Subjective measures


-0.5
-1
-1.5
p > 0.56
Triano

0
Symbio

0.5
Diva

male voice in quiet
female voice in quiet
own voice
male voice in impulse
noise
piano music
quiet room
10 NH
10 HI
1
Claro

1.5
Canta
Round robin paired
comparison
preference test
Avg preference score

(Score of 8 shows consistent
preference)
Keidser et al.
10
8
6
Quiet room
4
2
0
-2
-4
HI
NH
-6
-8
-10
10
12
14
16
18
20
22
24
26
Internal noise re reference test gain (dB)
6
Normal hearing
listeners
Avg preference score
Avg preference score
National Acoustic Laboratories, Sydney, Australia
Male voice
Female voice
4
2
0
-2
-4
-6
0
2
4
6
8
Time delay (msec)
10
12
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Conclusion




No overall significant difference in sound quality
among devices
Devices with less internal noise preferred in
quiet surroundings
Normal-hearing listeners preferred devices with
less time delay (< 10 msec) for listening to
speech in quiet
Recommendation: fit devices with lower internal
noise and possibly with shorter processing time
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Signal processing and horizontal
localisation performance



Interaural time and level difference (ILD and
ITD) enables left/right discrimination while
monaural spectral cues above 4 kHz enables
front/rear discrimination
In linear devices tubing, transducers, and filters
cause time delays that may distort ITD, and
inadequate amplification above 4 kHz and
microphone location on BTE devices distort
spectral cues (e.g. Byrne et al., 1992)
In digital hearing aids the signal processing is
complex
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Signal processing and horizontal
localisation performance

Effect of
 Multi-channel
WDRC
 Noise reduction
 Directionality
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Response patterns
Linear amplification, pulsed pink noise
Normal hearing, pulsed pink noise
200
200
N = 12
150
100
100
50
50
0
0
-50
-50
-100
-100
-150
-150
-150
-100
-50
0
50
100
Presentation azimuth (degree)
150
200
Response azimuth (degree)
Response azimuth (degree)
-200
-200
N = 16
150
-200
-200
-150
-100
-50
0
50
100
Presentation azimuth (degree)
The hearing-impaired subjects produced front/rear confusions in 40% of
responses, presumably due to the microphone location on the BTE devices
150
200
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Effect of WDRC and NR
p = 0.09
Linear
WDRC
NR off
NR on
75
Total RMS error (degree)
Total RMS error (degree)
75
p = 0.24
70
65
60
55
70
65
60
55
50
50
pink noise
pink noise + activation noise
-The proportion of front/rear confusions was the same across the four
conditions.
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Effect of directionality
Front/rear confusions ignored
Omni pair
Cardioid pair
Fig8/Cardioid
Omni/Cardioid
Total RMS error (degree)
75
70
65
60
55
50
Modified RMS error (degree)
p = 0.007
30
28
26
24
Omni pair
Cardioid pair
Fig8/Cardioid
Omni/Cardioid
22
20
p = 0.00001
18
16
14
12
10
pink noise
Front/rear confusions reduced by
11%, on average, when fitted with
the cardioid pair and
omni/cardioid combination
pink noise
Microphone-mode mismatch
increased left/right errors.
Significant bias of perception
towards fig8 ear and omni ear
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Conclusion





Front/rear confusions are prominent in BTE
users
The impact of multi-channel WDRC and noise
reduction is considered unimportant
A cardioid characteristic can reduce front/rear
confusions
Microphone-mode mismatch increases left/right
confusions
Recommendation: Counsel BTE users and
clients fitted with adaptive directionality about
possible localisation problems
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Preference for direct or amplified
low-frequency sound



To date the most efficient solution to the
occlusion effect is a vent bore or open mould
that creates a direct sound path for lowfrequency sound
The direct sound path will reduce the potential
benefit from directional microphones and noise
reduction algorithms (Dillon, 2001)
Do hearing aid users prefer direct or amplified
sounds when features such as directionality and
noise reduction are enabled?
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Preference for direct or amplified
sound
Insertion gain in dB
25
20

N = 22

15
(0 dB, off)
(6 dB, off)
(12 dB, off)
(0 dB, on)
(6 dB, on)
(12 dB, on)
10
5
0
100
1000
Frequency in Hz
10000

HTL at 500 Hz ranged
from 12 to 65 dB HL
Fitted vent size
ranged from open to
1.5 mm (vent effects
were compensated
for in two responses)
Field evaluation of 4
weeks
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Preference for direct or amplified
sound
Features on
20
20
16
16
No of subjects
No of subjects
Features off
12
8
4
0
12
8
p = 0.03
4
0
6
12
dB insertion gain at 250 Hz
30 dB HL
34 dB HL
0
0
6
12
dB insertion gain at 250 Hz
28 dB HL
43 dB HL
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Preference for direct or amplified
sound
3.0
250 Hz
2.5
2.0
1.5
1.0
0.5
0.0
-0.5
DI measured on KEM
-1.0
-5
0
5
10
15
20
REIG measured on KEMAR (dB)
25
30
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Conclusion
Generally, there was a strong preference
for direct sound to amplified sound, even
with features such as directionality and
noise reduction enabled
 Recommendation: only compensate for
vent effects to reach a target insertion gain
of 3 dB or above rather than provide
sufficient gain to achieve effective
operation of hearing aid features

National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Compression parameters for
severe to profound hearing loss



Intuitively we would fit severe and profound
hearing loss with low compression thresholds
and high compression ratios in multiple channels;
a combination that has proved to adversely affect
speech recognition (Souza, 2002)
When fitted with moderate compression
parameters, people with severe to profound
hearing loss generally prefer WDRC to linear
amplification (Ringdahl et al., 2000; Barker et al., 2001)
What compression ratios in the low and high
frequencies are preferred by hearing aid users
with severe to profound hearing loss?
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Adaptive paired comparisons
HF CR
LF CR

1:1
1.8:1
3:1
X
1:1
X
X
1.8:1
X
X
3:1
X

X

21 subjects with
moderately severe
to severe-profound
hearing loss
3 weeks in the field
Diaries and exit
interview
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Preferred scheme
HF CR
LF CR
1:1
1.8:1
3:1
1:1
1
7
3
1.8:1
1
4
4
3:1
1
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Ranking order
(N varies from 5 to 21 across schemes)
4.5
4.5
p = 0.1
4.0
4.0
3.5
3.5
3.0
3.0
2.5
2.5
2.0
2.0
1.5
1.5
1.0
Ranking score (said prefere
Mean
Mean±SE
Mean±SD
0.5
Ranking score (diary ratings)
1.0
p = 0.004
Mean
Mean±SE
Mean±SD
0.5
1/1
1/1.8
1/3
1.8/1 1.8/1.8 1.8/3
3/1.8
3/3
Compression scheme (LF CR/HF CR)
1/1
1/1.8
1/3
1.8/1 1.8/1.8 1.8/3
3/1.8
3/3
Compression scheme (LF CR/HF CR)
On average, the schemes providing linear amplification in the low frequencies
were ranked highest
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Prediction
- Audiometric data?
- Onset of loss (congenital = 8 vs. acquired = 13)?
- Previous amplification experience (linear = 10 vs. non-linear = 11)?
1.1
1.1
1.0
1.0
0.9
0.9
0.8
0.8
?
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
Average 1/CR in HF ban
Average 1/CR in LF band
0.3
0.2
45
NO
NO
NO
0.3
50
55
60
65
70
75
80
Average HTL in LF band (dB HL)
85
90
0.2
60
65
70
75
80
85
90
95
Average HTL in HF band (dB HL)
100
105
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Conclusion
Predominant preference for compression
ratios between 1:1 and 2:1, with a
preference for a higher ratio in the high
than in the low frequencies
 Recommendation: Fit moderately severe
loss with (1.5:1, 2:1) and fit severeprofound loss with (1:1, 2:1). Fine-tuning is
essential!

National Acoustic Laboratories, Sydney, Australia
Keidser et al.
NAL-NL1 and gain adaptation


General belief that new hearing aid users prefer
less gain than experienced users and that new
users will acclimatise to more gain over time
No support in the literature (on average 2 dB
difference in preferred gain), but adaptation
managers are introduced in fitting software
(Convery et al., 2005)

Do new users prefer less gain than experienced
users overall, in the low, or in the high
frequencies?
National Acoustic Laboratories, Sydney, Australia
Study design in brief



60 new and 25
experienced (>3 years)
hearing aid users fitted
with the same type of
device
NAL-NL1, NAL-NL1 with
6 dB LF-cut, and NALNL1 with 6 dB HF-cut
Gain preference
measurements @ 3
weeks, 3 months, and 12
months
Keidser et al.
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Preferred - NAL-NL1 prescribed
4FA gain in dB
Gain preference @ 3 weeks
0
-1
-2
-3
-4
-5
-6
-7
-8
-9
-10
2.5 dB
Inexperienced Experienced
(N = 28)
(N = 12)
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Preferred - prescribed 4FA gain in dB
Gain preference with 4FA HTL
10
Experienced
Inexperienced
Linear (Inexperienced)
5
0
-5
r = -0.5, p = 0.006
-10
-15
0
10
20
30
40
50
60
70
4FA HTL in dB HL
Difference in gain preference reduced from 2.5 dB to 1.8 dB
National Acoustic Laboratories, Sydney, Australia
Gain preference over time
Preferred - NAL-NL1 prescribed
gain in dB
23 inexperienced hearing aid users
0
-1
-2
-3
-4
-5
-6
-7
-8
-9
-10
3 weeks
3 months
Keidser et al.
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Conclusion
Little evidence to support that new hearing
aid users prefer significantly less gain than
experienced users – at least when the
hearing loss ranges from mild to moderate
 Recommendation: don’t use adaptation
managers with NAL-NL1
 Data from this study will form part of the
revisions made in NAL-NL2

National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Summary



Avoid fitting digital aids with high level of internal
noise and possibly long processing delays
Be aware that BTE users may have great
difficulty discriminating between sounds coming
from the front and the rear and that adaptive
directionality may affect left/right discrimination
Remember that a microphone characteristic with
different sensitivity to sounds coming from front
and rear may enhance front/rear discrimination
in BTE users
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Summary continued



Don’t compensate for vent effects when fitting
clients with directionality and noise reduction
except to reach target gain of 3 dB or above
Don’t assume that a hearing aid user with a
severe/profound loss can’t benefit from WDRC
but fit this population with ratios in the range 1:1
to 2:1 and provide sufficient support to facilitate
fine-tuning
Don’t use adaptation managers when fitting new
hearing aid users with the NAL-NL1 target,
however, some fine-tuning may be needed
National Acoustic Laboratories, Sydney, Australia
Keidser et al.
Many thanks to





Tom Scheller from Bernafon,
Ole Dyrlund and Gary Gow from GN Resound,
Volkmar Hamacher, Kristin Rohrseitz, Joseff
Chalupper, and Matthias Froehlich from
Siemens Instruments, and
Anna O’Brien, Heidi Silberstein, Elizabeth
Convery, Lisa and David Hartley, Margot
McLelland, and Ingrid Yeend from NAL
Several audiologists from Australian Hearing
Thank you for
listening