Title: MCHAS
1NOISE REDUCTION STRATEGIES
Wave 4 SFR 17/05/04
2NOISE REDUCTION STRATEGIES
- Multi-microphone technology
- Noise Reduction Algorithms
- Expansion/Squelch
- Multi-channel compression i.e. BILL
3Whats The Problem?
- Interference from background noise represents the
problem - To improve speech perception in noise, hearing
aids need to improve the signal-to-noise ratio - Hearing impaired individuals require a greater
S/N ratio than normal hearing listeners - Hearing aids with omnidirectional microphones
amplify sounds from all directions in the
listeners environment equally
4Multi-Microphone Technology
- Directional amplification is the most effective
way to improve signal-to-noise-ratios in hearing
aids.
Dillon 2001
BKB sentences Moore, Johnson, Clark Pluvinage,
1992
5- Directional microphones enhance the hearing
instruments sensitivity to sounds from desired
direction over sound from other directions. - It is the only technology that can effectively
distinguish between desired and undesired speech
signals. It is therefore quite effective in so
called cocktail party situations.
6Directional Microphone
Important information
Adapted from Dilion 2001
Dillon 2001
7Dual Microphones
- Similar principle to single directional
microphone but has two microphones with 2
separate ports. - Important users have the option to switch between
omni-directional directional amplification.
8- The 2 microphones can be either
- 2 omni-directional ones whose outputs are coupled
with electronic time delay and subtraction
components to make them directional. When omni-
direction is needed just one of the microphones
is used. - 1 directional microphone (which has a front and
rear port) and the other an omni-directional
microphone. - The directional sensitivity of a microphone is
usually indicated on a polar plot.
9Cardioid Family
Super
90
60
120
150
30
Hyper
180
0
330
210
Figure-8
300
240
270
Dillon 2001
10(No Transcript)
11Microphone Arrays
- FIXED MICROPHONE ARRAYS
- Current hearing aids implement variable polar
patterns in one of two ways - The first method introduces circuitry that allows
either the user or the hearing aid programmer to
select the desired directional pattern
(presently, none of the aids on contract have
this option) - Such a system means that there will be the same
polar pattern in all situations
12Microphone Arrays
- ADAPTIVE MICROPHONE ARRAYS
- The second method is designed to adaptively
switch between polar patterns in response to the
listening environment. - In noisy environments adaptive instruments
automatically shift polar patterns in an attempt
to maximize S/N ratio in the presence of other
noise sources.
13Microphone Arrays
- It may appear that adaptive directional
processing has the potential to improve speech
recognition across noisy environments - However, little evidence to suggest more
beneficial than fixed directional processing - Ricketts Henry (2002) found advantages only in
specific environmental listening situations - Bentler et al (2003) found directional better
than omnidirectional but adaptive no better than
fixed.
14Microphone Arrays
- May be of more benefit when
- (Dillon, 2001)
- there is one dominant source of background noise
- When the noise source is very close
- When the aid wearer is in a large open area so
that echoes and reverberation do not affect the
operation of the microphone - However, further research required in this area!
15A Word of Caution
- Increased circuit noise in directional mode
- Less effective as speaker-listener distance
increases - Echoes/reverberation
- Microphone drift
- Dirt!
16Patient selection
- Everyone benefits from increased SNR.
- Younger children must be able to hear sounds from
all directions so must have access to
omni-directional microphone. - Directional depends on maturity
- Those with good low frequency hearing will hear
low frequency noise through vents therefore the
benefit of the directional microphone will be
reduced.
17Phonak Aero 211, 311 413 AZ
- Provides omnidirectional and fixed directional
(supercardiod) microphone performance. Either
microphone can be implemented in specific
programs.
18Danalogic 163D
- Provides omnidirectional and fixed directional
(hypercardiod) microphone performance.
Accessible in all programs.
19Danalogic 283D
- Provides omnidirectional and fixed directional
(hypercardiod) microphone performance.
Accessible in all programs.
20Prisma 2M 2D SP
- Directional microphone accessible in both
channels. - Fixed supercardiod
- Self calibrating to avoid microphone drift (uses
2 omni-directional mics coupled together)
21Oticon Spirit II D
- Directional in program 2 only, and can be
disabled. - Fixed supercardiod array.
- In program 2, can access different degrees of LF
cut.
22Noise Reduction Algorithms
- 2 types of noise reduction systems
- Modulation Detection Synchrony detection
- We will concentrate on modulation detection as it
is far more common at present.
23Modulation Detection
- Underlying algorithms may vary between
manufacturers but overall aim is to provide
improved listener comfort and possibly improved
speech intelligibility in background noise. - Common to all algorithms is this need to identify
which signal is noise and which is speech.
24Modulation Detection
- A speech/non speech detector analyses
fluctuations in signal amplitude. - Speech noise envelopes fluctuate in a well
characterised manner - Speech modulations tend to be slow and have big
amplitude fluctuations. - Noise modulations tend to be more constant with
rapid and smaller fluctuations.
25Envelope of speech noise
26Modulation Detection
- The precise way in which the hearing aid
processes the two signals will differ depending
on the algorithm. - One common approach is to
- Estimate when the speech signal is present in
each channel. - If the amplitude envelope at a channels output
is characteristic of speech the gain in that
channel remains unaltered.
27Modulation Detection
- If the envelope is relatively constant then the
signal is assumed to be noise and the gain for
that channel is reduced. - If more noise is detected than speech at a
certain channel, then the gain will also be
reduced.
28- Important points to note
- If the unwanted signal has similar fluctuations
to speech, then it is unlikely to be attenuated. - Best results occur with steady state noise that
has a narrow frequency band, or is of low
frequency (to reduce the risk of upward spread of
masking) - If any other constant amplitude signals (e.g.
pure tones) are identified then they will also be
treated as a noise signal and be attenuated.
This means that pure tones should not be used to
either set or test the hearing aid, unless the
noise reduction feature has been turned off.
29Summary
- Hearing aids with omni-directional microphones
fail to provide the required S/N improvement - Perhaps not so blatantly obvious
- This is true for ALL hearing aids with
omni-directional microphones... - including digital hearing aids with adaptive
noise reduction or speech recognition
algorithms.
30Whats Wrong with this Picture? (Starkey
laboratories, 2003)
31Modulation DetectionAdvantages and Limitations
- ADVANTAGE Channel specific gain reductions to
ensure comfort whenever noise is present - LIMITATION Cannot distinguish between noise only
versus speech plus noise - Tryanski Pogash
- (AAA 2204)
32Synchrony Detection
- Assumption you can find speech versus noise but
you cannot separate them - Designed to look for the unique structure of
speech (energy) - Maintains full response whenever speech is
present.only goes into comfort mode when speech
is no longer present - Tyranski Pogash
- (AAA 2004)
33Synchrony DetectionAdvantages Limitations
- ADVANTAGE Can protect audibility of speech
whenever speech is present - LIMITATION Anytime speech is present there will
be no comfort-based changes, even if noise
present and high level - Tyranski Pogash
- (AAA 2004)
34The Future of Noise Reduction?
- One company have recently launched a system that
combines the two??
35Conclusions
- Very little research on NR systems and none at
all with the paediatric population - Alcantara et al (2003) found there to be no
improvement in sound quality or listening comfort
with noise reduction turned on - Boymans Dreschler (2000) found no extra benefit
with NR Directional when compared to
directional alone - Therefore, research published to date would
suggest that directional microphone systems lead
to a better improvement in the S/N ratio
36Phonak Aero 211AZ, 311AZ Supero 413 digital
Noise Canceler (dNC)
- Light indicates up to 3 dB gain reduction
- Moderate indicates up to 6 dB gain reduction
- Strong indicates up to 12 dB gain reduction
- OFF deactivates the dNC in any program
37Danalogic speech/ noise analysis in 14 frequency
bands
normal gain
38Danalogic 283D Noise Reduction
- Uses modulation or level fluctuations to identify
speech from noise.
Noise Reduction either ON or OFF
39Prisma 2M 2D SP
- Can be deactivated.
- Reduction of gain can be altered to min12 dB,
med18 dB and max24 dB. Defaults to min. - Gain reduction is prioritised in channel 1 4.
C1 C2 C3 C4 Off - - - - Min min - - min Med med m
in min med Max max med med max
40Starkey Strata 312
- Off 0dB reduction
- Min Up to a 10dB channel gain reduction
(default) - Max Up to a 20dB channel gain reduction
41Oticon Spirit II
- No control over settings.
- Identifies speech noise signals based on
modulation index. - If signal characteristic of noise the release
time will be delayed by up to 4 times. If
characteristic of speech there will be no change
in release time.
42Expansion/Squelch
- Expansion or squelch occurs when the compression
ratio is less than 11, therefore gain decreases
as the input level decreases. - It is useful for decreasing the audibility of
very low level sounds. Good as long as ALL
inaudible sounds are unwanted e.g. microphone
internal noise, computer hum.
43Expansion Squelch
- By removing low level electronic noises it can be
thought of as a low level noise reduction
strategy. - Important to distinguish between true noise
reduction algorithms and squelch. - NHS DSP Aids with expansion/squelch
- Phonak Aero, Oticon Spirit II, Starkey Strata.
44Multi-channel compression- BILL
- BILL compression circuits were designed to
improve speech intelligibility in background
noise by providing more gain for soft low
frequency sounds than for loud low frequency
sounds. - Based on the premise that background noise is
predominantly low frequency, quite loud and hence
would be attenuated while high frequency sounds
would remain unaffected.
45Multi-channel compression- BILL
- Any aid that has
- Multi-channel compression
- Adjustable kneepoints
- Adjustable compression ratios
- . could be configured to a BILL response.
46Questions?