Supplemental Digital Text Entry #2: Advances in Cochlear Implant Technology:

hopefulrebelAI and Robotics

Nov 24, 2013 (3 years and 8 months ago)


Supplemental Digital Text Entry #2:
Advances in Cochlear Implant Technology

Spahr and colleagues (6,25) examined the importance of input signal processing and showed that
use of a wider input dynamic range (IDR) (e.g., > 30 dB) along with automatic gain

(AGC) compression resulted in better speech recognition in competing noise. Specifically,
Advanced Bionics recipients using a sound processor with a 60 dB IDR and AGC achieved
significantly better speech recognition in noise when compared to Coch
lear recipients using the
Nucleus® ESPrit™ 3G sound processor, which possesses a 30 dB instantaneous input dynamic
range (IIDR). The IIDR is defined as
the range of short
term, instantaneous fluctuations that are
mapped without compression (or other types

of attenuation, such as Autosensitivity™ Control
[ASC]) into the recipient’s electrical dynamic range at a given point in time. The IIDR is in
contrast to the IDR, which is the total range of input levels the processor can analyze without
clipping or sig
nificant distortion and reports an overall IDR of 75 dB for the Nucleus®
Freedom™ and Cochlear® Nucleus® 5 CP810 Sound Processors. Cochlear’s use of two
distinctive terms, IDR and IIDR, is intended to acknowledge the fact that at any given point in
a 40 dB range of inputs is presented into the recipient’s electrical dynamic range, but as the
overall input level to the microphone changes, ASC, Adaptive Dynamic Range Optimization
(ADRO®), and compression allow for the processor to code a 75 dB range of

(approximately 20 to 95 dB SPL) without significant distortion.

The poorer performance for Cochlear recipients in the Spahr et al. (2007) study highlights the
importance of proper implementation of input processing within a cochlear implant sound
processor. When the Nucleus ESPrit 3G is set at default settings, all acoustic inputs between 35
and 65 dB SPL are mapped into the recipient’s electrical dynamic range. Inputs below 35 dB
SPL are mapped below the recipient’s electrical threshold, while i
nputs exceeding 65 dB SPL are
subjected to infinite compression. As a result, in noisy conditions where the speech and noise
are likely to exceed 65 dB SPL, the limiting compression results in an embedded speech signal
within the ongoing background nois
e. A manual or automatic reduction in microphone
sensitivity, as recommended by the manufacturer for use in noisy situations, would have allowed
for better speech recognition in noise, but the subjects were tested in noise at inappropriate sound

settings (i.e., microphone sensitivity settings that are only recommended for quiet

Subsequent research studies have examined performance in noise for Cochlear recipients using
sound processor settings recommended for use in noise. Wolfe et
al (8) assessed whether use of
the Cochlear input signal processing strategy, Autosensitivity Control (ASC), would result in an
improvement in speech recognition in noise. In this study, sentence recognition in the presence
of classroom noise (7) was meas
ured at multiple, fixed SNRs for a group of 12 adult cochlear
implant recipients in an ASC disabled and enabled condition. Wolfe and colleagues (8) found
speech recognition in noise improved by as much as 50% for the group when ASC was enabled.

Wolfe a
nd colleagues (26) also examined the effect of ASC on speech recognition in noise for a
group of pediatric cochlear implant recipients. A sentence
noise test was administered to 11
children (4 to 11 years
age) using Nucleus Freedom and Nucleus 5 sou
nd processors to
determine the difference in SRT performance between the ASC disabled and enabled conditions.
The researchers reported a significant mean improvement of 3.4 dB in the ASC
enabled relative
to the ASC
disabled condition.

The significant i
mprovement in speech recognition in noise with use of ASC is not surprising.
ASC essentially functions as a slow
acting AGC in noisy conditions. When the ambient noise
level at the sound processor microphone exceeds 57 dB SPL, the ASC algorithm decreases

sensitivity of the sound processor microphone in an attempt to maximize the distance between
the peaks of the speech signal and the ongoing competing noise. During this process, the
deleterious effects of limiting compression at 65 dB SPL, observed b
y Spahr et al (6,25), are
reduced or eliminated, because the reduction in microphone sensitivity prevents the input from
being subjected to hard limiting.

Another form of Cochlear input signal processing, Adaptive Dynamic Range Optimization
(ADRO), also

aims to improve performance in difficult listening situations. ADRO adaptively
and constantly adjusts the gain in each of the channels of the sound processor based on an
analysis conducted across three features of the input signal: the average input leve
l, the level of
the background noise, and the level of the loudest sounds. To maximize speech recognition,
gains are increased in channels that possess a favorable SNR or include low
level speech.
Conversely, to normalize loudness and prevent discomfort,

gains are decreased in channels with
intensity sounds.

Clinical studies evaluating speech recognition in noise with ADRO have produced mixed results.
Dawson et al (27) studied speech recognition for a group of 15 children using ADRO with the
eus® SPrint™ body
worn sound processor. Speech recognition in quiet and in noise was
significantly better with use of the ADRO program as compared to the children’s standard
program, which did not include pre
processing. Additionally, everyday sounds wer
e comfortable
with ADRO (27). In contrast, James and colleagues (28) evaluated speech recognition in quiet
and in noise for a group of nine adults using ADRO with the SPrint sound processor and found
that ADRO improved recognition of low
level speech in q
uiet relative to the standard program,
but provided no significant improvement in speech recognition in noise. James et al (28) did
report a significant improvement in perceptual sound quality with ADRO compared to the
conventional program.

In addit
ion to signal processing advances, it is well
known that the use of directional microphone
technology provides substantial improvement in speech recognition in noise. In fact, hearing aid
research has shown that, other than the use of a personal FM system
, directional microphone
technology is the most effective means to improve speech recognition in noise (29,30). There is,
however, a paucity of published studies examining the potential benefit of commercially
available directional technology for cochlear

implant recipients.

Spriet et al (31) evaluated speech recognition in noise for a limited number of Nucleus Freedom
recipients (n = 5) using the Beam™ input processing strategy. In this study, the Nucleus
Freedom sound processor had two microphones: (
a) a single
cartridge, dual
port directional
microphone, which provided a relatively modest amount of attenuation of signals arriving from
the rear hemisphere for default everyday settings and (b) a separate omni
directional microphone
that could be used i
n conjunction with the directional microphone for beamforming. The Beam
processing utilized digital signal processing and summation of the signals from the
directional and omni
directional microphones of the Freedom sound processor. The goal of the
daptive beamforming is to focus on signals arriving from in front of the user and to
automatically change the position of the null at the sides or behind the user to provide maximal
attenuation for noise signals moving throughout the rear hemisphere. Spri
et and colleagues (31)
determined the SRT for word and sentence materials in the presence of both competing speech
babble and steady
state noise fixed at 55 and 65 dB SPL. The speech signals of interest were
presented from a loudspeaker located at 0 degre
es in the horizontal azimuth, and the noise was
presented from (a) one loudspeaker directly to the side of the implanted ear and (b) three
loudspeakers at 90, 180, and 270 degrees. When the noise was presented to the side of the user,
Beam provided a mean

SRT improvement of 7.2 (55 dB SPL) and up to 15.9 dB (65 dB SPL)
across the two types of noise when compared to the conventional directional mode. For the 90,
180, 270 noise condition, Beam provided a mean SRT improvement of 1.5 to 5.3 dB at a 55 dB

presentation level and 6.5 and 11.6 dB at a 65 dB SPL presentation level for the steady
state and babble noise, respectively. The authors did not mention whether the pre
strategies ADRO and/or ASC were used in conjunction with Beam. At the t
ime of the study,
simultaneous use of the aforementioned Cochlear input processing strategies was not
commercially available. As a result, Beam pre
processing was most likely used in isolation (i.e.,
without ASC and/or ADRO).

Recently, Gifford and Revit

(32) evaluated speech recognition in quiet and in noise for 20 adult
recipients using the Nucleus Freedom sound processor with a variety of pre
processing schemes.
They used the Revitronix R
Space system to present restaurant noise at an overall level of

dBA fro
m seven loudspeakers separated by 45 degrees from 45 to 315 degrees in the horizontal
azimuth. Two lists of sentences from the Hearing in Noise Test (HINT) were presented from a
loudspeaker located at 0 degrees azimuth, and the SRT was determined for each

subject. Also,
speech recognition in quiet was assessed with a full list (50 words) of CNC (consonant
consonant) monosyllabic words presented at 60 dBA. Performance for both measures was
assessed in a variety of conditions: (a) ADRO only, which
was the manufacturer
“Everyday” program for use in most realistic situations with the Freedom processor, (b)
ASC+ADRO, which was the manufacturer
recommended “Noise” program for the Freedom for
use in listening situations in which competing noi
se was present, and (c) BEAM+ASC+ADRO,
which the manufacturer referred to as the “Focus” program for Freedom users to use when the
competing noise is present and the signal of interest arrives from the front. Gifford and Revit
(32) reported that the “Nois
e” program (i.e., ASC+ADRO) improved the mean SRT by 2.5 dB
compared the “Everyday” program (ADRO alone), and the “Focus” program
(ASC+ADRO+Beam) improved the mean SRT by an additional 3.6 dB when compared to the
“Noise” program.