|
Digital
Hearing Aids - the Way of the Future
Steven D. Clements, School of Human Communication Disorders, Dalhousie University, Halifax, Nova Scotia, Canada This paper was published in proceedings from "Acoustic Week in Canada 1991" - CAA Conference, Edmonton, Alberta, Canada, October 7 - 10, 1991 Copyright 1991-2014, Digital Recordings. All Rights Reserved.
Introduction
First developments in digital processing of speech sound were done
in 1960's in Bell Laboratories [29]. In recent years, the use of digital
components in the design of hearing aids (HA's) is fast becoming a
standard rather than an exception. Various HA's on the market, that
utilize digital circuits, have proven to provide increased flexibility
and efficiency in both fitting and hearing aid evaluation [1,2,25].
Although the availability of digital signal processing (DSP)
techniques that are applied to the incoming signal do not vary much
in the commercially available units, it is evident that research in DSP
is the most promising area in future hearing aid development.
Already new generation of HA's brings higher levels of satisfaction
from the end user-hearing impaired [2,17,24,26].
Signal Processing can be viewed as any manipulation of a signal that
alters its characteristics; whether extracting, enhancing or otherwise
modifying said information [3,4,30]. These changes are conducted
in order to help the hearing aid wearer to better discriminate between
speech and noise; essentially then to increase signal to noise ratio
(S/N). To achieve these results the consumer has available to him/her
two basic approaches: automatic/adaptive digitally controlled
analogue systems, and digitally programmable HA's. It should be
noted that however complex a HA may seem, generally five
approaches to signal processing are in use: variable gain,
equalization, compression, limiting, and steady state noise reduction
[8]. Following is a description of the major types of DSP on the
market.
This form of DSP; decreasing certain frequencies, while trying to
leave the so called speech frequencies untouched, has been shown
to increase speech intelligibility by 15% in the presence of low
frequency noise [3], but so far gives no increased benefit to the HA
wearer while in the presence of competing noise of similar broadband
spectra, such as cafeteria noise [7].
The Argosy Manhattan II Circuit
automatically alters its frequency
response as a result of continuous sampling of sounds in the
environment. As the noise SPL increases, high frequency gain
decreases and low frequency output also decreases. Adjustments to
the low frequency potentiometer provides up to 40 dB of gain
reduction at 500 Hz [4].
The Siemens 283 changes its overall
frequency response to
compensate for the loss in speech intelligibility caused by low
frequency noise. The incoming signal is divided into two channels,
the low frequency; up to 800 Hz which contains a compression
circuit and the linear high frequency channel, which contains an
adjustable high-pass filter from 800-1600 Hz [3].
From Intellitech the
Zeta Noise Blocker II, a digital microchip
integrated into a hearing aid circuit, samples incoming signals and
analyzes the rate of frequency change, becoming active when the
presence of noise is detected and applying digitally controlled
attenuation by four analog filters. This circuit has been shown to be
more effective in increasing S/N in the presence of high frequency
competition [1,9].
Although the K-AMP is an amplifier,
it does adaptively affect the
linearity of a signal, offering approximately 25 dB maximum gain
for sounds below 40 dB SPL, with gain gradually reduced to 0 dB
as input level increase. The K-AMP will operate to 110 - 115 dB
SPL input without distortion, and is noted as only amplifying quiet
sounds. Loud transient sounds that represent a problem to many HA
users are passed without amplification. These transients cause the
Telex Communication Adaptive Compression Circuit °, included
in the K-AMP, to quickly drop to 0 dB, with a recovery time almost
as fast; resulting in little affect on the ongoing gain [5,6].
Digitally programmable HA's process sound in the analogue domain,
but are controlled by digital circuitry. They have the ability to be
reprogrammed by an external unit, thus allowing quick caparisons of
different settings in order to determine patient's preferences [1,10].
Paired-comparison techniques can be used with these types of HA's
to improve the fitting process [22].
The Audiotone System 2000
(from Dalberg "The Dolphin
SystemTM") was the first programmable unit available on the market
(1988). Its programming features include, maximum output, gain,
high frequency cut (25 dB at 5 kHz), low frequency cut (30 dB at 500
Hz), input compression and frequency dependent output limiting
[1,11].
The Maico/Bernafon PHOX
(programmable hearing operation
system) allows the programmability of the following features: gain,
high and low frequency cut, slope of the high and low frequency
response, a high frequency emphasis filter (cuts gain below 1500
Hz), a 3 kHz peak to emulate the canal resonance, option X a patient-
activated noise reduction system, which cuts high and low frequency
gain an additional 18 dB per octave, and an automatic
gain control
(AGC) input compression circuit [1,11,12,13,25].
Siemens Triton 3000 is a
three channel compression HA. Ability
is provided to program the gain in each channel (up to 36 dB), the
AGC in each channel as well as two crossover points (300-1400 Hz
& 1-5 kHz) with a minimum bandwidth of 1 octave and maximum
of 3 octaves [1,11,25].
Resound's Personal Hearing System (PHS)
offers a programmable
two channel compression unit,with a compression range of 3:1 - 1:1,
and an ultrasonic remote control for reprogramming. Programmable
crossover frequencies for the high and low frequency bands are
between 400-4000 Hz, with gain variable from 0 - 40 dB for each
band. Fitting information can be stored in memory cartridge, as well
as transferred to a PC. ReSound also incorporates an input
compression limiter to control maximum output [1,11,14,25].
3M Corporation offers the
Memory Mate, a programmable two
channel compression HA. Programmable functions include; overall
gain, dispenser-adjustable crossover frequency between 500-4000
Hz, and an eight memory client selected RAM. The RAM stores
different frequency responses, for different environments [1,11,25].
The Widex Quattro allows
for programmability of gain, maximum
power output, output compression on/off, low and high-cut filters,
plus an additional low-cut filter, the inverse presbycusis adaption
filter. The HA also contains four memory choices selectable from a
small FM remote control, which also operates as a programming unit,
with the insertion of a programming key [1,11,15,16,25].
The Ensoniq Sound Selector HA
has a programmable thirteen band
equalizer that is adjustable in 1 dB increments up to 40 dB in each
band and up to 60 dB overall gain. The thirteen bands are divided as
such: lowest two in 1 1/2 and 1 octave bands respectively, the next
three are 1/2 octaves, and the following eight are 1/3 octave bands.
This is noted to generate a smooth response virtually free of
distortion caused by resonance peaks. Also standard is 2:1
frequency-independent input compression and a directional
microphone [1,11,17,18,19,25].
The Nicolet Phoenix, the
only fully digital HA, has been taken off
the market. Based upon the fitting, different frequency response
characteristics, and a noise reduction algorithm were programmed
(by the manufacturer) into the HA. The user was then able to select
one of three "modes", by selecting one of three buttons on the aid,
each set up for different listening environments. The
Phoenix VP
(variable processor) has five frequency response curves programmed
into the first button, five degrees of noise reduction in the second,
and the third button set to reverse the effect of the first two. The
Phoenix Plus offers an additional
15 dB of gain [1,24].
Other programmable hearing instruments are offered by:
Audioscience (2-channel),
Beltone Electronics Corp. (1-channel),
GN Danavox (3-channel),
Philips Hearing Instruments (3-channel),
Rexton Inc. (1-channel)
and Starkey Laboratories Inc.
(1-channel) [25].
In the past decade digital HA research and development has
resulted in a number of improvements in clarity and S/N ratio (in
certain environments). Most of current ASP (automatic signal
processing) devices are still limited however by the number of
frequency bands; thus are generally only effective in the presence of
low frequency competition. All but Ensoniq have limited their
approach to a few distinct processing channels . We should note
when considering masking by noise, it is well accepted that the
frequency range of the human auditory system is divided into 24
discreet channels, or critical bands (for each ear) [20] and each
cochlear neuron responds to a narrow range of frequency stimuli
[21]. It would seem appropriate then to at least separate the frequency
range into like divisions. Philips has recently developed the Digital
Compact Cassette (DCC), which utilizes 32-band digital processing
for perceptual coding of audio signals [27]. This technology could be
very easily adopted to make 32-band HA's. However, DCC will not
be accepted as a new recording format, since recordable optical discs
will be released soon to the consumer market.
We still lack sufficient understanding as to the nature of many
hearing impairment problems and their relationship to one another
[2,5,29].Generally hearing problems are divided into: conductive,
cochlear, eighth nerve and central nervous system disorders [2].
However the number of the distinctive disorders may reach easily
into the hundreds [2], and each may require a specific signal
processing algorithm [2,24]. Further research is still needed to
explore the usefulness of compression systems [23]. Information
theory can be applied to calculate inherent channel capacity for the
ear [28]. On the basis of this theory analysis of a hearing impaired
communication channel could be performed and the most appropriate
information coding obtained.
Digital Signal Processing workstations (for example NeXT
computer) should be used to perform further psychoacoustic tests in
order to learn more about the human auditory system. Also a Digital
Master HA can be simulated on these types of computers and used to
design and check different DSP strategies to be used in HA's.
The complexity of processing which is needed to address many
hearing disorders requires highly sophisticated signal processing.
DSP offers substantial improvements over analog techniques [30]
along with unmatched flexibility and precision to adopt the
processing to individual requirements of each patient [2]. Also the
paired-comparison judgment technique may be used more effectively
with this technology for precise HA fitting [22].
DSP should complement rather then substitute for signal
processing which is performed in the auditory system (in other words
it should be transparent when not needed). This will allow the best
signal processor so far- the human brain - to extract information most
efficiently [18].
The following DSP techniques could be used in future HA's :
arbitrary filtering and frequency shaping, arbitrary gain (as function
of frequency and signal amplitude), frequency shifting, feedback
control, noise reduction (various techniques), peak clipping or
limiting etc [24,30]. Also multichannel, parallel processing can be
done with DSP improving speed and sophistication of sound
processing. "Smart HA's" with adaptive algorithms and performing
logical operations can be build around DSP technology to further
improve HA's capabilities.
In our opinion DSP is still an underexplored technology in the
area of HA's, but this may change in the near future with
anticipated benefits to the hearing impaired.
HA's however sophisticated never would be a panacea for
hearing impairment. Hearing impairment reduces information
channel capacity (from outside world to auditory system) and this
can't be restored with a hearing aid. HA's can only help to better
utilize the remaining information channel capacity.
References
Terms of Use | Return Policy | Privacy Policy
No part of the information provided on this www page may be reproduced for any purpose, in any form, without prior written approval.
|