Introduction to NALNL1

1
Introduction to NAL-NL1

• Harvey Dillon, Gitte Keidser, Denis Byrne,
  Richard Katsch, and Teresa Ching
• National Acoustic Laboratories and
• Co-operative Research Centre for
• Cochlear Implant and Hearing Aid Innovation

2
Proposed fitting procedures for non-linear
hearing aids

• LGOB, IHAFF, DSLi/o, FIG6, and ScalAdapt
• Loudness Normalisation
• Aim To make a sound at any frequency and at any
  input level just as loud for the impaired ear as
  it is for normal ears

3
Why doubt loudness normalisation?

• Dont account for greater loudness of low
  frequencies of speech

4
International Long-term Average Speech Spectrum
5
Loudness perception
6
Loudness perception
7
Why doubt loudness normalisation?

• Dont account for greater loudness of low
  frequencies of speech
• Dont account for reduced effectiveness of
  audibility at frequencies of greatest hearing
  loss

8
Effect of hearing loss
1
Normal hearing

Effective audibility
hearing impaired
0
30
0
Sensation level (dB)
9
Why doubt loudness normalisation?

• Doesnt account for greater loudness of low
  frequencies of speech
• Doesnt account for reduced effectiveness of
  audibility at frequencies of greatest hearing
  loss
• Research shows that normalisation is not correct
  for average speech inputs

10
Rationale of NAL-NL1 procedure

• Maximise speech intelligibility for any input
  level
• For each input level, make overall loudness of
  speech equal to, or less than, normal

11
Predicting speech intelligibility
Level distortion factor

• SII ? Wi.Ii.Li

1
0
73
30
1/3 octave SPL
Threshold
Frequency
12
Predicting speech intelligibility
Level distortion factor

• SII ? Wi.Ii.Li
• SII ? Weffi.Ii.Li

1
0
73
30
1/3 octave SPL
Threshold
Frequency
13
Predicting loudness

• Loudness model of Moore and Glasberg

Calculate loudness per band
Sum across bands
Filtering into auditory bands
External middle ear
14
Predicting loudness

• Loudness model of Moore and Glasberg

Calculate loudness per band
Sum across bands
Filtering into auditory bands
External middle ear
Allowance for hearing loss
15
The derivation

• Choose an audiogram
• Choose an input level
• Calculate normal loudness
• Juggle gains until SII is a maximum
  and loudness is
  normal or less

16
The derivation (cont)

• Repeat for several input levels,
• Repeat for many audiograms,
• Fit an equation to these optimum gains.

17
NAL-NL1 insertion gain

• IG at each frequency, depends on
• HTL at that frequency
• 3 freq. Average HTL (500, 1000, 2000 Hz)
• Slope from 500-2000 Hz
• Overall speech input level
• Thus, gain at any frequency depends on HTL at
  several frequencies, as for NAL-RP

18
Observations on prescriptions

• Loudness is approximately constant across
  frequency

80
65
Specific Loudness
50
Frequency
19
Observations on prescriptions

• Loudness is approximately constant across
  frequency
• Gain-frequency response does not normalise
  loudness across frequency

80
65
Specific Loudness
50
Frequency
Loudness
500
3k
20
Observations (cont)

• At medium levels, gain and frequency response
  agree with NAL-RP

21
Access to NAL-NL1

• NAL-NL1 Stand-alone Clinician software
  (www.nal.gov.au)
• Hearing aid manufacturers software and analysing
  equipment (Bernafon, Frye Electronics,
  Intrason, Rion, Siemens, Sonic Innovations,
  Starkey, Unitron, Viennatone)

22
Evaluation of NAL-NL1
23
Fitting Wide Dynamic Range Compression Hearing
Aids

• Gitte Keidser, Harvey Dillon,
• and Frances Grant
• National Acoustic Laboratories and
• Co-operative Research Centre for
• Cochlear Implant and Hearing Aid Innovation

24
Fitting WDRC Devices

• Fitting rationale
• Number of channels
• OSPL90 prescription
• Conductive component
• Verifying the response
• Multiple memories

25
Fitting Rationale

• Loudness Normalisation or Speech Intelligibility
  Maximisation?

26
Test design

• Prescription procedures IHAFF and NAL-NL1 (two
  channels)
• Twenty-four subjects flat and steeply sloping
  losses
• Laboratory tests paired comparison and speech
  recognition (BKB sentences)
• Field test (4-6 weeks) satisfaction rating and
  description of performance in individually
  selected listening situations

27
(No Transcript)
28
Prescribed CR
LF
HF
29
Paired Comparison
30
(No Transcript)
31
(No Transcript)
32
Speech recognition test

33
Field test
34
(No Transcript)
35
Recommendation

• Use NAL-NL1 as starting point
• If NAL-NL1 is very different from LN then NAL-NL1
  is preferred
• If difference from LN is small then the preferred
  response shape is reached during fine-tuning

36
How many Channels?
37
Multi-channel compression
Advantages

• Improved audibility (especially for sloping
  losses)
• Improved SNR when noise is dominant in a
  restricted range of frequencies

38
Multi-channel compression

• Past research based on loudness normalisation
  procedures
• LN procedures prescribe high CRs and low CTs
• High CR and low CT negatively affect speech
  intelligibility in multi-channel schemes

39
Test design

• NAL-NL1 implemented in one, two and four channels
  
• Twenty-four subjects flat and steeply sloping
  losses
• Laboratory tests paired comparison and speech
  recognition (BKB sentences)
• Field test (4-6 weeks) satisfaction rating and
  description of performance in individually
  selected listening situations

40
Paired comparison
41
Input/Output curves (all)
f 300 Hz
f 800 Hz
f 2000 Hz
f 4000 Hz
42
Speech recognition test(in traffic noise)
43
Field test
44
Discriminating factors
45
Recommendation

• For clients with sloping losses use at least two
  compression channels
• For clients with flat losses the number of
  compression channels is probably unimportant

46
Questions
47
OSPL90 Prescription

• Not to high - not too low

48
OSPL90 in WDRC devices

• OSPL90 is less critical because gain decreases
  when input level increases

21 comp.
Linear
49
OSPL90 in WDRC devices

• Use the same OSPL90 prescription as for linear
  devices
• Most based on avoiding discomfort, so set just
  below LDL
• POGO 4 dB below 3FA LDL
• Cox (1985) 100 dB SPL 1/2 HTL
• NAL-SSPL halfway between minimum and maximum
  output limit

50
OSPL90 in WDRC devices

• In most cases the OSPL90 level produced is below
  what is prescribed
• Is OSPL90 too low? - likely to depend on the
  prescribed compression characteristic

51
Multi Channel Hearing Aids

• Limiting may occur on the wideband signal after
  the channels are recombined, or independently in
  each channel before the channels are recombined
• If OSPL90 can be independently adjusted in each
  channel - allow for effects of power and loudness
  summation

52
Multi Channel Hearing Aids
Total Leq 110 dB SPL
Total Leq 115 dB SPL
53
Conductive Component
54
Prescribed Gain

• Linear devices apply additional gain equal to
  75 of the conductive component
• It is possible that for non-linear devices, 100
  of the conductive component can be applied to
  low-level signals ( increased CR)

55
Prescribed OSPL90

• If OSPL90 is prescribed on the basis of measured
  UCL use procedure as is
• If using NAL-SSPL 1) Prescribe OSPL90 for
  sensorineural part, 2) increase OSPL90 by 87.5
  of conductive component (Min. Limit increases by
  75 and Max. Limit increases by 100 )

56
Verifying the Fitting
57
Real-ear measurements?
YES

• Real-ear gain at different input levels (e.g. 50,
  65, and 80 dB SPL)

• Real-ear gain at one input level (e.g. 65 dB SPL)
  and input/output curves (one for each compression
  channel)

Use a broadband stimulus for measuring gain
58
Subjective evaluation

• Check that speech presented at 50, 65, and 80 dB
  SPL are rated about soft, comfortable, and
  loud
• If speech at 80 dB SPL is rated loud the OSPL90
  is not too low
• If other high-level sounds are acceptable the
  OSPL90 is not too high

59
Multiple memories

• Candidacy and selecting alternative amplification
  characteristics

60
Candidacy
Criteria

• Difficulty hearing in acoustically diverse
  conditions
• Can understand and manage multi-memory concept
• Variation in low frequency real-ear gain of at
  least 5 dB achievable
• Average high frequency hearing loss greater than
  55 dB HL

61
Client needs help hearing in acoustically diverse
listening situations? (interview or COSI)
62
Response variation

• Prescriptive responses
• known to be effective
• based on audiogram
• Use NAL-RP as reference to compensate for the
  individual hearing loss

63
Response selection
LISTENING CONDITION
PREFERRED RESPONSE

• Speech in quietSpeech in reverberation
  NAL responseSpeech in babble-noise__________
  ______________________________________________
• Ease of understanding a dialogue High
  frequency comp.(distant talkers, soft voice)
  (CR 41, CT 55 dB)___________________
  _____________________________________
• Ease of understanding speech in Steeper
  responselow frequency background noise
  (re. NAL response, 4-6 dB)_______________________
  _________________________________
• Reduce annoyance of low Low
  frequency comp.frequency background noise
  (CR 21, CT 55 dB)
  __________________________________________________
  ______
• Reduce annoyance of high Flatter
  responsefrequency background noise
  (re. NAL response, 4-6 dB)

64
(No Transcript)
65
Linear variation
NAL-RP as base response
NAL-RP
LF
NAL-RP
HF
66
Linear variation
NAL-RP as base response
NAL-RP
LF
Flatter
NAL-RP
HF
Flatter
67
Linear variation
NAL-RP as base response
NAL-RP
LF
Flatter
Steeper
NAL-RP
HF
Steeper
Flatter
68
Linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
Flatter
Steeper
NAL-RP
NAL-NL1
HF
HF
Steeper
Flatter
69
Linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
Flatter
Steeper
NAL-RP
NAL-NL1
HF
HF
Steeper
Flatter
70
Linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
Flatter
Steeper
NAL-RP
NAL-NL1
HF
HF
Steeper
Flatter
71
Linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
Flatter
Steeper
Increase or decrease gain, and keep CR and CT
the same
NAL-RP
NAL-NL1
HF
HF
Steeper
Flatter
72
Non-linear variation
NAL-RP as base response
NAL-RP
LF
NAL-RP
HF
73
Non-linear variation
NAL-RP as base response
NAL-RP
LF
LF comp.
NAL-RP
HF
74
Non-linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
LF comp.
NAL-RP
NAL-NL1
HF
HF
75
Non-linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
LF comp.
NAL-RP
NAL-NL1
HF
HF
76
Non-linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
LF comp.
Increase CR and decrease CT in low frequency
band
NAL-RP
NAL-NL1
HF
HF
77
Non-linear variation
NAL-RP as base response
NAL-RP
LF
NAL-RP
HF
78
Non-linear variation
NAL-RP as base response
NAL-RP
LF
LF comp.
NAL-RP
HF
HF comp.
79
Non-linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
LF comp.
NAL-RP
NAL-NL1
HF
HF
HF comp.
80
Non-linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
LF comp.
NAL-RP
NAL-NL1
HF
HF
HF comp.
81
Non-linear variation
NAL-RP as base response
NAL-NL1 as base response
NAL-RP
NAL-NL1
LF
LF
LF comp.
Increase CR in both bands. Decrease CT in LF
band and increase CT in HF band
NAL-RP
NAL-NL1
HF
HF
HF comp.
82
Notice

• Verify fitting using insertion gain
• Ensure that client can tell the programs apart
• Follow-up 3-4 weeks later - adjust or try
  alternative, if needed
• We dont know the best combination of response
  variation and hearing aid facility

83
The end
84
NAL-SSPL (Dillon Storey)

• Defined min. limit to avoid saturation (re. 75 dB
  SPL speech amplified according to NAL-RP)
• Defined max. limit to avoid discomfort (re.
  measured LDL values)
• Prescription midway between the two limits
• Evaluation Suitable for 80 of subjects, but
  always verify the OSPL90 setting!

85
NAL-SSPL Procedure
86
Multi Channel Hearing Aids
Number of channels Reduction (dB) 1
0 2 5 3 7 4 9 5 10
87
Otosclerosis

• A possible change in bone conduction thresholds
  because of stiffening or fixation of the stapes
• Decrease bone conduction thresholds before
  calculating gain for either the s/n or the
  conductive component

88
Prescribed Gain

• For conductive HLoss HTL, MCL and LDL are all
  increased by the same amount, however, do not
  prescribe 1 dB of gain for every 1 dB of HLoss
• Not researched, but apply additional gain equal
  to 75 of the conductive component
• It is possible that for non-linear devices, 100
  of the conductive component can be applied to
  low-level signals ( increased CR)