What CI designers can learn from hearing aid designers

1
What CI designers can learn from hearing aid
designers
Harvey Dillon National Acoustic Laboratories
(NAL), CRC for Cochlear Implant and Hearing Aid
Innovation
Asilomar August, 2003
NAL
2
These slides can be down-loaded from the NAL
web-site

• www.NAL.gov.au
• from August 27
• Research_at_NAL.GOV.AU

3
The future
H. Dillon NAL, CRC for CI and HAI
4
Similarities of CI and HA design problem

• Input signal corrupted by noise and reverberation
• Dynamic range from threshold to discomfort less
  than normal (dB and number of jnds)
• Frequency resolution less than normal
• Constraints on signal processing due to power
  consumption constraints (diminishing constraint)
• Miniaturisation and ease of operation (hence need
  for automation)
• Probably long-term impact of use on the auditory
  system

H. Dillon NAL, CRC for CI and HAI
5
Handling reduced dynamic range
Wide dynamic range compression
6
Wide dynamic range compression
Intense
Non-linear
Moderate
Weak
Normal
H. Dillon NAL, CRC for CI and HAI
7
Wide dynamic range compression
(a.k.a. non-linear amplification)

• For intense sounds
• ? More comfort ..
• For medium sounds
• ??? No advantage .
• For weak sounds
• ??? Intelligibiity increase ..

• Automatic operation
• convenient for anyone .
• essential for those with manipulation difficulty.

H. Dillon NAL, CRC for CI and HAI
8
Wide dynamic range compression
Intelligibility benefit
Moore, Johnson, Clark Pluvinage, 1992
H. Dillon NAL, CRC for CI and HAI
9
But what do people want?
H. Dillon NAL, CRC for CI and HAI
10
Should wide dynamic range compression be fast or
slow?
H. Dillon NAL, CRC for CI and HAI
11
Compressed envelopes
Input
Ts
H. Dillon NAL, CRC for CI and HAI
12
Should wide dynamic range compression be fast or
slow?

• Some results and conclusions from Stuart
  Gatehouse ..

H. Dillon NAL, CRC for CI and HAI
13
S. Gatehouse
14
S. Gatehouse
15
S. Gatehouse
16
S Gatehouse
17
Extreme compression in CIs
(Speak / Ace)
Source Franck, Xu Pfingst JARO (2002)
0449-59
18
Extreme compression in CIs
(Speak / Ace)
Source Franck, Xu Pfingst JARO (2002)
0449-59
19
Lessons Compression

• Use it,
• but how?
• Available loudness range
• to represent inter-syllabic differences, or
  inter-environment differences?
• Optimal time-constant unclear,
• but combinations likely to be better than either
  fast or slow alone
• Optimal compression threshold unclear
• but need experimental evaluation
• Assess intelligibility and quality

H. Dillon NAL, CRC for CI and HAI
20
Using two ears
Bilateral devices
21
Bilateral or binaural amplification
Speech
H. Dillon NAL, CRC for CI and HAI
22
Demonstration of head diffraction effects
Noise
Speech
H. Dillon NAL, CRC for CI and HAI
23
Terminology (and clear thinking)
Separate speech intelligibility benefits of
bilateral devices into those provided by

• Redundancy
• the same information presented twice
• a.k.a. binaural summation, and diotic benefit
• Squelch
• combining equally good SNRs by using phase/timing
  differences
• Head shadow
• taking advantage of superior SNR at one ear

H. Dillon NAL, CRC for CI and HAI
24
Redundancy
Woof
Source Ching et al. (2003)
H. Dillon NAL, CRC for CI and HAI
25
Squelch ( redundancy)
Speech
Noise
Noise700
Ching et al. (2003)
H. Dillon NAL, CRC for CI and HAI
26
Head shadow ( squelch, redundancy)
Noise
Speech
Ching et al. (2001)
H. Dillon NAL, CRC for CI and HAI
27
Bilateral implants Better than ????
H. Dillon NAL, CRC for CI and HAI
28
Bilateral or binaural amplification

• Clearer in quiet
• Clearer in noise
• Better localisation
• Greatest benefit for people with moderate to
  profound hearing loss
• Average 5 dB improvement in signal to noise
  ratio
• Avoiding auditory deprivation

H. Dillon NAL, CRC for CI and HAI
29
Lessons Binaural

• If its aidable, aid it (for hearing aids)!
• The head shadow advantage can be huge
• Two microphones and intelligence to use it
• Be clear about which binaural advantage(s) is
  being measured
• CI HA is the benchmark, not unilateral CI

H. Dillon NAL, CRC for CI and HAI
30
Improving signal to noise ratio
Directional Microphones
31
The biggest problem noise
32
Effect of decreasing noise
BKB sentences Moore, Johnson, Clark Pluvinage,
1992
H. Dillon NAL, CRC for CI and HAI
33
Directional microphones
H. Dillon NAL, CRC for CI and HAI
34
Conventional Directional microphone
Front port tube
Rear port tube
Diaphragm
Acoustic damper
Hearing aid case
Source Dillon (2001) Hearing Aids
35
Polar diagrams
Front
H. Dillon NAL, CRC for CI and HAI
36
Directional microphone
H. Dillon NAL, CRC for CI and HAI
Front
37
Directional microphone
Front
H. Dillon NAL, CRC for CI and HAI
38
Correcting microphone drift
Front
?
-

?
T
Adaptor
Output
Source Dillon (2001) Hearing Aids
39
Omni on head
Cardioid on head
Source Dillon (2001) Hearing Aids
40
Directivity index

• Variations
• 2-D metric
• derived 3-D metric
• measured 3-D metric

Front
Directivity index
AIDI DI weighted across frequency by
Articulation Index weights
H. Dillon NAL, CRC for CI and HAI
41
Directivity Index
H. Dillon NAL, CRC for CI and HAI
42
Subtractive array
-
Directional at low and mid freq -
Directivity limited to 6 dB - Low-frequency
cut - Higher internal noise

?
T
Output
Source Dillon (2001) Hearing Aids
43
Source Dillon (2001) Hearing Aids
44
Super-directional microphone
Second-order subtraction
d
Front
-

?
T
H. Dillon NAL, CRC for CI and HAI
45
45 mm
Source Raicevich (1996)
H. Dillon NAL, CRC for CI and HAI
46
Directivity index - hand-held
H. Dillon NAL, CRC for CI and HAI
47
Two-sided array Griffiths Jim adaptive noise
reduction
Right

?


?
-
Left
Source Dillon (2001) Hearing Aids
48
Microphone array benefit
Source Hoffman et al (1994)
H. Dillon NAL, CRC for CI and HAI
49
Directional mics - when and whereQuestion
3 What are the characteristics of the everyday
listening situations that favor either the OMNI
or the DIR microphone mode?

• Answer Environmental characteristics that
  favor
• Omnidirectional Directional
  
  
  
• Talker located other than front Talker located
  in front
• Background noise absent, very Moderate
  background noise
  soft, or very loud present
• In cars or out-of-doors Talker distance
  lt 20 feet
• Smaller rooms Larger Rooms
• 
  People as background noise

B Walden
50
Lessons Directional microphones

• They work - users know when and where
• They need switching - manual or automatic
• Still great potential for improvement

H. Dillon NAL, CRC for CI and HAI
51
Maximising comfort in noise
Adaptive noise suppression
52
Adaptive noise suppression
Comfort, not intelligibility
H. Dillon NAL, CRC for CI and HAI
53
H. Dillon NAL, CRC for CI and HAI
54
Noise Reduction
Signal
Input Spectral Level
Noise
Frequency
Gain
Signal
Output Spectral Level
Noise
H. Dillon NAL, CRC for CI and HAI
55
Detecting SNR

• 1. Histogram approach

H. Dillon NAL, CRC for CI and HAI
56
Steady HF-noise
Impulsive HF-noise
Babble-noise
Keidser Grant
Traffic-noise
Speech in quiet
Less LF/more HF gain in dB
More LF/less HF gain in dB
NAL-response
57
Environmentally sensitive hearing aids
Telecoil input
Magnetic signals
Directional mic
Wideband response
?
Low-cut
Low-cut
Wind
H. Dillon NAL, CRC for CI and HAI
58
Adaptive speech detection
H. Dillon NAL, CRC for CI and HAI
59
Lessons Adaptive noise suppression

• Comfort advantage, usually no intelligibility
  advantage
• Dont just measure intelligibility - comfort also
  important
• Probably applicable to implants (e.g. ADRO)
• To some extent, Most CI mapping schemes already
  do it.

H. Dillon NAL, CRC for CI and HAI
60
Miscellaneous
61
Wireless / Direct audio input
H. Dillon NAL, CRC for CI and HAI
62
H. Dillon NAL, CRC for CI and HAI
63
H. Dillon NAL, CRC for CI and HAI
64
Multi-memory
Different programs for different situations
H. Dillon NAL, CRC for CI and HAI
65
Transposition
Intensity
1000
250
4000
Frequency
H. Dillon NAL, CRC for CI and HAI
Source Dillon (2001) Hearing Aids
66
Spectral Enhancement
Intensity
1000
250
4000
Frequency
Inferrred excitation pattern
Optimistically hoped-for excitation pattern
H. Dillon NAL, CRC for CI and HAI
67
Spectral enhancement
/ata/
Original
Enhanced
Time (seconds)
Source Dillon (2001) Hearing Aids
68
Transient emphasis
H. Dillon NAL, CRC for CI and HAI
69
Lesson

• What seems like it should work doesnt always!
• e.g. 1. Spectral enhancement
• e.g. 2. Normalising loudness

H. Dillon NAL, CRC for CI and HAI
70
Thats all Folks
71
Web-site address WWW.NAL.gov.au(including
instructions for Teresa Chings procedure on
adjusting hearing aids to complement cochlear
implants)

• Research_at_NAL.GOV.AU