Research and Technology
Paper
Benefits of Adaptive FM Systems
on Speech Recognition in Noise
for Listeners Who Use Hearing Aids
Linda Thibodeau
University of Texas at Dallas, Callier Center
for Communication Disorders
Purpose: To compare the benefits of adaptive
FM and fixed FM systems through measurement of speech recognition in noise with
adults and students in clinical and real-world
settings.
Method: Five adults and 5 students with
moderate-to-severe hearing loss completed
objective and subjective speech recognition in
noise measures with the 2 types of FM processing. Sentence recognition was evaluated
in a classroom for 5 competing noise levels
ranging from 54 to 80 dBA while the FM microphone was positioned 6 in. from the signal
loudspeaker to receive input at 84 dB SPL.
The subjective measures included 2 classroom
activities and 6 auditory lessons in a noisy,
public aquarium.
T
he use of FM systems can provide significant improvements in speech recognition in noisy environments for
persons with normal hearing as well as persons with
impaired hearing who wear hearing aids (Boothroyd, 2004;
M. S. Lewis, Crandell, Valente, & Enrietto Horn, 2004) or
cochlear implants (Schafer & Thibodeau, 2003, 2004, 2006).
By having the talker wear a microphone that transmits the
speech signal to a receiver worn by the listener, the signal-tonoise ratio can be dramatically increased. There are various
options now for FM transmitters and receivers, as shown
in Tables 1 and 2, that include location and directivity of the
microphone, number and synchronization of channels with
receivers, and programmable features. Selection of these features is often dependent on user preference; however, some
require verification to illustrate benefit (Thibodeau, 2004).
Early research with FM systems involved comparisons
of various arrangements for delivering the FM signal to the
Disclosure Statement
The information listed in this article was presented at Phonak training seminars
and at the July 2009 annual meeting of the Educational Audiology Association
(New Orleans, LA). Phonak provided financial support for this study.
36
Results: On the objective measures, adaptive
FM processing resulted in significantly better
speech recognition in noise than fixed FM processing for 68- and 73-dBA noise levels. On the
subjective measures, all individuals preferred
adaptive over fixed processing for half of the activities. Adaptive processing was also preferred by
most (8–9) individuals for the remaining 4 activities.
Conclusion: The adaptive FM processing
resulted in significant improvements at the higher
noise levels and was preferred by the majority
of participants in most of the conditions.
Key Words: hearing loss, FM systems,
speech recognition
user, such as neckloop or direct-audio-input (DAI) connections. Hawkins (1984) found that the DAI arrangement with
directional microphones on the personal behind-the-ear hearing aids provided the greatest benefit for speech recognition
in noise compared with neckloop arrangements. In subsequent
studies, electroacoustic testing with various FM coupling
arrangements was performed and helped to explain why the
DAI connections were superior to neckloop arrangements
(Thibodeau, 1990; Thibodeau, McCaffrey, & Abrahamson,
1988; Thibodeau & Saucedo, 1990). Electroacoustic characteristics for the combined hearing aid and FM arrangements
were more similar to the characteristics of the hearing aid
alone when using DAI arrangements compared with neckloop.
Furthermore, undesirable variations in output up to 20 dB were
reported with changes in positioning of the neckloop relative
to the t-coil in the hearing aid (Thibodeau et al., 1988). Electroacoustic analysis with more recent neckloop and DAI connections to hearing aids with advanced circuitry confirmed
the benefits of DAI over neckloop arrangements. Schafer,
Thibodeau, Whalen, and Overson (2007) reported that use of
FM with neckloops resulted in reduced low-frequency output
relative to the frequency response of the hearing aid alone.
American Journal of Audiology • Vol. 19 • 36–45 • June 2010 • A American Speech-Language-Hearing Association
Downloaded from aja.asha.org on June 14, 2010
Table 1. Available options for FM transmitters (adapted from American Academy of
Audiology [AAA], 2008).
Feature
Variations
Microphone
Options
Location
External (lapel)
External (head boom, cheek boom)
Internal (lavalier)
Conference
Omnidirectional
Directional fixed
Directional user select
Single fixed
Multichannel selectable
72–76 MHz
169–176 MHz
216–217 MHz
Direct (user changes channel)
Automatic (channel changes when in
close proximity to receiver)
Trimpot
Dial
Reduce number to those frequently used
NiCad only
NiCad and alkaline
Lithium-ion polymer
User access
Childproof, but with user access
Manufacturer only
Low battery
FM transmission
Audio input jack
Interface jack for data transfer, programming
and charging
Option to turn on
Type
Channel
Number
Frequency
Synthesis
Controls
AGC
Volume
Channels
Type
Programmable
Batteries
Replacement
Indicator lights
On front
Accessory jacks
Secondary transmission
Note.
Bluetooth
AGC = automatic gain control.
They quantified the difference between the output of the hearing aid alone and the hearing-aid-plus-FM system as the FM
advantage. According to professional guidelines (American
Academy of Audiology, 2008; American Speech-LanguageHearing Association, 2002), the desired FM advantage for
optimum speech recognition in noise is 10 dB. However,
Schafer et al. reported that hearing aids with advanced compression circuits often did not show the desired 10-dB FM
advantage when measured with an FM system via a DAI
connection. They concluded that because the testing protocol
Table 2. Available options for FM receivers (adapted from AAA, 2008).
Feature
Variations
Options
Location
Self-contained
Interface with hearing aid
Channel
Number
FM receiver within BTE case
Design integrated receiver
Receiver attaches to BTE via audio shoe
Single fixed
Multichannel selectable
72–76 MHz
169–176 MHz
216–217 MHz
Default channel
Start-up at default channel or last used channel
Channel set for scan
Adjust level of FM signal regarding microphone
signal
Adjust gain and output of BTE
Continuous or interrupted
FM, FM + HA, HA
FM, FM + HA, HA
Zinc air
Frequency
Programmable
Frequencies
FM /HA ratio
Switches
Batteries
Note.
Gain/output
Indicator lights
Programmable
Mechanical
Type
BTE = behind the ear; HA = hearing aid.
Thibodeau: Adaptive FM
Downloaded from aja.asha.org on June 14, 2010
37
involved sequential measurements, first the hearing aid followed by the hearing-aid-plus-FM, the compression features
confounded the observation of the true FM advantage. This
measurement protocol, in fact, was not like the real-world
settings where there are simultaneous inputs to the hearing aid
and FM microphones. These results suggested the need for
more behavioral testing to document benefits of the FM advantage settings, particularly in real-world arrangements.
Only one report of speech recognition as a function of FM
advantage setting has been published. In an attempt to mimic
real-world listening conditions, D. Lewis and Eiten (2004)
recorded speech in noise at different FM advantage settings.
They conducted a survey with audiologists who listened
to the recordings of speech at the different FM advantage settings. Listeners preferred the greater FM advantage for
audibility of the talker’s voice, but this resulted in reduced
audibility of self and of other voices. The optimal FM advantage setting may actually vary depending on the communication situation and the ambient noise.
One recently developed feature in the host of FM arrangement options is known as adaptive FM advantage (AFMA).
Fixed FM systems allow the talker’s voice to be presented
at a constant level above the level of the signal from the local
microphone on the user’s hearing aids. Typically, this is the
recommended 10-dB FM advantage; however, in programmable FM receivers, this can be adjusted if some prefer to
hear more or less of the conversation around them. When the
background noise is high, one might prefer to increase the
signal-to-noise ratio to 15 dB or to switch to an FM-only
setting. With the new adaptive FM systems, the system automatically adjusts the output of the FM receiver on the basis
of the ambient noise. If the transmitter detects that the ambient noise exceeds 57 dB SPL, a signal is sent to the FM
receiver to increase the gain of the FM receiver and therefore
increase the FM advantage to a maximum of 24 dB. There
is no additional change in the signal processing other than the
increase in the FM advantage. The signal sent to the receiver
is a digital code delivered with the FM signal at a different
inaudible audio frequency. If the background noise level decreases below 57 dB SPL, the system will return to the 10-dB
FM advantage. The AFMA system involves a new transmitter
with a directional lapel microphone or a boom microphone
and new FM receivers that can decode and react to the digital
signals sent by the transmitter.
Success with AFMA processing has been reported for
listeners using two types of cochlear implants: Advanced
Bionics implants and Cochlear Corporation implants (Wolfe
et al., 2009). Using the Phonak Inspiro FM Transmitter and
the MLxi (AFMA) or MLxS (fixed) FM Receivers with their
implants, listeners repeated sentences from the Hearing in
Noise Test (HINT; Nilsson, Soli, & Sullivan, 1994) in background noise levels of 55, 65, 70, and 75 dBA. There was
a significant improvement with the AFMA processing compared with the fixed processing for both types of implants.
The purpose of the present study was to compare benefits of
AFMA and fixed FM advantage systems through objective
and subjective measures in adults and students who use hearing aids in clinical and real-world settings.
Method
Participants
Five adults (20–55 years of age) and five students
(11–15 years of age) with moderate-to-severe hearing loss
who wore binaural behind-the-ear hearing aids were invited
to participate. Descriptive information is provided in Table 3.
All participants were experienced FM users and agreed to
use the system over a 1-week period. Within the student
Table 3. Descriptive information for the 10 participants.
Frequency (Hz)
Participant
Age
( years)
1
14
2
12
3
15
4
11
5
42
6
30
7
12
8
49
9
55
10
20
Note.
Ear
250
500
1K
2K
4K
8K
PTA
Unaided
speech
Aided
speech
Hearing aids
DPAI
R
L
R
L
R
L
R
L
R
L
R
L
R
L
R
L
R
L
R
L
60
65
75
65
65
60
45
40
55
65
50
60
60
60
15
15
65
70
60
80
70
70
80
65
70
55
50
50
65
75
65
60
65
70
25
30
54
70
70
80
70
70
80
70
70
70
65
55
75
90
65
65
70
65
45
25
60
65
80
85
65
65
75
65
65
65
55
50
70
85
70
75
75
70
70
70
45
50
80
80
50
50
70
55
50
55
45
45
65
75
70
65
65
70
105
105
40
50
70
70
50
45
75
60
40
50
n/a
n/a
85
80
85
65
80
80
NR
NR
50
50
NR
95
52
52
78
67
68
63
57
52
65
81
67
67
70
67
42
45
70
70
70
80
100
96
100
88
64
76
92
88
96
88
88
92
92
92
42
45
84
88
44
56
88
Widex B2
No
100
Rexton Sensation
No
92
Siemens Music D
No
100
Siemens Music D
No
100
Phonak Savia 311
Yes
100
Widex Senso Diva
No
92
Phonak NovoForte 4
Yes
88
Oticon Synchro 2
Yes
100
Phonak eXtra 311
Yes
80
Phonak eXtra 411
Yes
PTA = pure-tone average; DPAI = Direct Programmable Audio Interface; R = right; L = left.
38 American Journal of Audiology • Vol. 19 • 36–45 • June 2010
Downloaded from aja.asha.org on June 14, 2010
group, there were two sets of siblings. Each participant demonstrated a receptive vocabulary age equivalent score greater
than 9 years as measured by the Peabody Picture Vocabulary Test—Revised (Dunn & Dunn, 1981). All were paid
$100 for their participation in the week-long project.
Equipment
Prototype body-worn FM receivers and a special version
of adaptive FM transmitters in SmartLink housings were
used. The FM signal was delivered to the participant’s personal hearing aid via a DAI connection. The body-worn unit
had two switches (A and B) that allowed reception of the
fixed and AFMA processing. The switches were counterbalanced so that for half of the participants, the A setting was
the adaptive FM, and for the other half, the B setting was
the adaptive FM. To control for possible bias, participants
were blinded to the selection of the two systems.
Electroacoustic analysis using an FP 40 Hearing Aid Test
Box of the hearing aid alone and the hearing aid connected
to the FM system revealed distortion < 10% and reception of
signals via both the hearing aid and the FM microphones.
Speech recognition testing was conducted in a large classroom (28 × 24 feet) with four loudspeakers placed in the corners to present classroom noise and one at the front to deliver
the speech. The noise was delivered from a CD player and
was routed through an amplifier to the four loudspeakers
placed in the room as shown in Figure 1. The ambient noise
level in the classroom was 52 dBA.
Procedure
The evaluation of the two FM processing schemes included
objective and subjective measures and was conducted in two
individual sessions and one group session. Objective measures
of speech recognition in noise were obtained in a controlled
classroom. Subjective measures of benefit were obtained in
Figure 1. Classroom arrangement for speech recognition testing.
both the classroom and in a real-world setting over a 1-week
period. Prior to testing, the participants were allowed to select
a comfortable listening level in quiet through the AFMA
processing while the examiner spoke from another room at
an average conversational level and adjusted the volume on
the transmitter. The examiner spoke for 10 s at the initial
setting of “0” and asked the participant to judge the level as
“too soft, comfortable, or too loud.” The level was increased
to the next step, and the 10-s spoken sample and comfort
questions were repeated until a judgment of “too loud” was
reached. The comfortable listening level was designated as the
setting judged as “comfortable” on two ascending sequences.
This setting varied from the minimum of “0” to the maximum
of “8” and remained there for both the AFMA and the fixed
FM processing conditions for each participant. There was no
relationship between the type of aid or hearing aid gain and the
selected comfort setting. This procedure allowed for potential
equalization of differences across hearing aids that may have
resulted from variations in the impedance levels via the DAI
connection.
Objective Speech Recognition Measures
To compare the two types of signal processing (AFMA vs.
fixed FM) across five noise conditions in which differences
may potentially be minimal, it was necessary to use sensitive
speech recognition materials and to not duplicate sentences.
The two sets of materials were extracted from the HINT and
the Speech Perception in Noise Test (SPIN; Bilger, Nuetzel,
Rabinowitz, & Rzeczkowski, 1984). The HINT and SPIN
sentences were presented via the CD player at 84 dBA as
measured at the FM transmitter microphone that was placed
6 in. in front of the signal loudspeaker at the front of the room.
Ten HINT and SPIN sentences were presented for each noise
condition. The first five low- and five high-predictability SPIN
sentences were presented from each 50-sentence list. Because
the SPIN recording has a randomized order of high- and
low-predictability sentences, 10–12 sentences per condition
were required to meet the presentation criteria of five of each
type.
Classroom noise created from mixing recordings taken
in four elementary classrooms was presented from the four
competition loudspeakers at five noise levels: 54, 63, 68,
73, and 80 dBA as measured at the participant’s location for
the HINT (Schafer & Thibodeau, 2006). This noise was edited
to reduce peak intensities and resulted in minimum and
maximum root-mean-square values that differed by only
2.95 dB. The same noise was used for the SPIN; however,
to complete the data collection efficiently on this final,
fourth consecutive day of participation, testing was performed
at only three levels: 63, 68, and 73 dBA. The lowest level,
54 dBA, was omitted because differences between the processing schemes were expected to be minimal at this noise
level on the basis of the initial testing with the HINT. The
highest level, 80 dBA, was also omitted because the task was
very difficult and because participants were often not able to
repeat HINT sentences at that level. Two CD players were
used to present the noise so that the noise from the front left
and back right loudspeakers was uncorrelated with the noise
from the front right and back left loudspeakers.
Thibodeau: Adaptive FM
Downloaded from aja.asha.org on June 14, 2010
39
Within each measure, the order of noise levels and type of
FM processing were counterbalanced. One examiner sat next
to the participant to control the delivery of the classroom
noise while another sat next to the signal loudspeaker to
control the delivery of the primary signal. These examiners
and an additional assistant recorded the participant’s verbal
responses following each presentation. A condition was
aborted if none of the words could be repeated after the first
five sentences. This occurred for five participants at the
73-dBA noise level and for 10 of the participants at the
80-dBA noise level when in the fixed FM condition.
Figure 2. Sample rating card.
Subjective Speech Recognition Ratings
Two classroom activities and six lessons in a public aquarium were included in the subjective evaluation and are described in Table 4. During each activity, participants listened
to Settings A and B and then rated their listening experience
on index cards, as shown in Figure 2. For convenience, the
eight cards were carried in a plastic index card case that
could be worn around the neck when in the classroom and
when moving around the aquarium.
The classroom activities included a paper-folding task and
a group game. While recorded classroom noise was played
at 65 dBA, instructions were presented by an experimenter
who wore the FM microphone and who spoke at a natural
conversational level. The participants were seated in desks
with folding tops in three rows, and they faced the loudspeaker. Starting with the receivers on Setting A, 10 instructions
were given for folding with time for participants to complete
each task. Similarly, 10 instructions were given for folding
while participants listened in Setting B. Then they all completed the subjective rating card.
For the group game, the same recorded classroom noise
was played while playing the game to guess phrases. The
transmitter microphone was passed around the circle for each
person to give clues regarding a word he or she saw on a
prompter device that made a ticking sound. When the word
was correctly identified, the prompter was passed on to the
next person. The object was to not be holding the prompter
when the ticking stopped and a buzzer sounded. All participants started the game with their receivers set to Setting A and
switched to Setting B after 10 min of play. Then they completed another subjective rating card.
At the aquarium, participants listened to a tour guide who
wore the FM transmitter at each of six stations. After 2 min
of instruction at a station, they were asked to switch the
receiver setting from A to B. Then another 2 min of instruction regarding the station was provided by the tour guide.
Before moving to the next station, they completed the
subjective rating card. The ambient noise at each station
Table 4. Sequence of subjective listening activities.
No.
Activity
Location
1
Origami
Classroom
2
Catch phrase
game
Classroom
3
Tour guide
stations
Aquarium
Note.
% selecting
AFMA
Task
Participants were seated in classroom
chairs with folding desktops. One
research assistant wore the transmitter
and gave instructions to fold paper to
make a flower while classroom noise
was played in surrounding speakers at
65 dBA.
Participants were seated in chairs in a
large circle. The transmitter microphone
was passed around as each participant
gave clues regarding a word or phrase
for others to guess while the same
noise used in No. 1 was presented.
A trained tour guide wore the transmitter
and provided information regarding
6 aquarium exhibits while the ambient
noise levels ranged from 70 to 95 dBA.
The participants stood around the guide.
Stations
40-foot waterfall
Sloth in rainforest
Reptile exhibit
Manatee feeding
Aquatic tunnel
Coral and seahorses
AFMA = adaptive FM advantage.
40 American Journal of Audiology • Vol. 19 • 36–45 • June 2010
Downloaded from aja.asha.org on June 14, 2010
% selecting
fixed
100
0
100
0
90
100
100
80
80
90
10
0
0
20
20
10
varied depending on the animals and nearby visitors and
ranged from 70 to 90 dBA.
were three scoring methods used to attempt to thoroughly
capture differences in the processing methods: first word
correct, total number of words correct, and whole sentences
correct. The improvements in speech recognition with AFMA
versus fixed processing ranged from 0% at the 54-dBA noise
level for whole sentence correct to 50% at the 73-dBA noise
level for first word correct. It is not surprising that the differences
were minimal at the lowest noise level because the AFMA
processing is not engaged until the background noise level
exceeds 57 dB SPL.
For statistical analysis, the final noise level was omitted
because many participants were unable to complete the
test. Main effects of processing type (AFMA vs. fixed)
and noise level (54, 63, 68, and 73 dBA) were significant
for each scoring method: first word, F(1, 72) = 16.62,
p < .0001, and F(3, 72) = 21.34, p < .0001, respectively; total
words, F(1, 72) = 22.61, p < .0001, and F(3, 72) = 25.59,
p < .0001, respectively; and total sentences, F(1, 72) = 16.64,
p < .0001, and F(3, 72) = 18.02, p < .0001, respectively.
There was also a significant Processing Type × Noise Level
interaction for each scoring method: first word, F(3, 72) = 3.95,
p < .01; total words, F(3, 72) = 6.57, p < .0001; and total
sentences, F(3, 72) = 3.07, p < .05. Follow-up analysis with
Bonferonni adjustments for multiple t tests (Sankoh et al.,
1997) revealed that AFMA processing was significantly
better than fixed for the 68- and 73-dBA noise levels but
not for the lower noise levels of 54 and 63 dBA for each
scoring method. As shown in Table 5, for each scoring method,
the largest increase in speech recognition for AFMA over fixed
processing occurred at the 73-dBA noise level: 50% for first
word correct, 49.7% for total number of words correct, and
36% for whole sentence correct.
The performance on the SPIN is shown in Table 6 and
Figure 4. The percentage of increase in recognition for SPIN
sentences for AFMA versus fixed processing ranged from
Sequence of Events
The FM experience occurred over 1 week beginning on a
Friday with confirmation of the electroacoustic response of
the hearing aid and the DAI connection. On a Monday or
Tuesday, the FM prototype was introduced, and the HINT
was performed. The participants were then instructed to use
the device at home as much as possible to increase their
listening opportunities. On Wednesday, all 10 participants
returned for a full day of activities, which included the two
communication activities in the classroom and six educational activities while touring an aquarium (as shown in
Table 4). On Thursday or Friday, the SPIN was administered
in the same classroom setting used for the HINT.
Results
Objective Measures
The results for the HINT and SPIN are presented in Tables 5
and 6 and in Figures 3 and 4, respectively. Despite the age
range of the participants, 11–55 years, there was no evidence
that this was a significant factor: Correlations of scores with
age were nonsignificant and ranged from r(60) = .0, p < .05, to
r(60) = .21, p < .05. Therefore, all analyses were performed
with all 10 participants as a group. All scores were arc-sin
transformed prior to statistical analysis to guard against potential effects of unequal variances associated with the percentage of correct scores. Bonferonni adjustments for multiple
t tests were made when post hoc analyses were conducted
(Sankoh, Huque, & Dubey, 1997).
The average percentages of correct scores for the HINT for
each noise level are shown in Table 5 and Figure 3. There
Table 5. Mean percentage of correct scores and confidence intervals for HINT words and sentences by noise level.
Scoring category
First word correct
Total no. of words correct
Whole sentence correct
Note.
Noise
level
(dBA)
M score
AFMA
(%)
M score
fixed
(%)
M AFMA to
fixed difference
(%)
Post hoc t values
for AFMA
versus fixed
Confidence
interval
No. of participants
exceeding confidence
interval
54
63
68
73
80
54
63
68
73
80
54
63
68
73
80
90.0
79.0
77.0
60.0
25.6
85.2
76.0
68.9
55.0
22.7
70.0
54.0
52.0
36.0
11.1
88.0
71.0
53.0
10.0
UCTa
89.3
66.0
42.1
5.3
UCTa
70.0
37.0
19.0
0.0
UCTa
2.0
8.0
24.0
50.0
20.7
–4.1
10.0
26.8
49.7
22.3
0.0
17.0
33.0
36.0
11.1
0.00
1.39
5.53*
6.54*
UCTa
0.19
1.53
7.31*
7.09*
UCTa
–0.12
2.35
6.68*
4.23*
UCTa
8.9
15.6
17.2
21.9
UCTa
11.1
18.3
17.7
20.8
UCTa
19.1
23.2
21.0
23.4
UCTa
4
3
6
8
UCTa
2
4
8
7
UCTa
2
2
8
4
UCTa
HINT = Hearing in Noise Test; UCT = unable to complete test.
a
Testing at the highest noise level was discontinued when none of the words were intelligible after 5 sentences; therefore, the confidence interval
was not calculated.
*p < .01 (df = 9).
Thibodeau: Adaptive FM
Downloaded from aja.asha.org on June 14, 2010
41
Table 6. Mean percentage of correct scores and confidence intervals for high- and low-predictability SPIN sentences by noise level.
Scoring category
High-predictability score
Low-predictability score
Total score
Note.
Noise
level
(dBA)
M score
AFMA
(%)
M score
fixed
(%)
M AFMA to
fixed difference
(%)
Post hoc t values
for AFMA
versus fixed
Confidence
interval
No. of participants
exceeding confidence
interval
63
68
73
63
68
73
63
68
73
68.0
78.0
52.0
58.0
56.0
36.0
63.0
67.0
44.0
64.0
44.0
10.0
46.0
24.0
6.0
55.0
34.0
8.0
4.0
34.0
42.0
12.0
32.0
30.0
8.0
33.0
36.0
0.23
3.73*
2.98*
0.98
3.11*
6.12*
0.77
4.43*
4.58*
20.5
24.1
25.4
20.7
21.1
16.2
18.8
16.2
18.2
3
6
7
2
5
9
3
7
7
SPIN = Speech Perception in Noise Test.
*p < .01 (df = 9).
4% at the 63-dBA noise level to 42% at the 73-dBA noise
level for high-predictability sentences. Likewise, the improvement for the low-predictability sentences increased with
these noise levels from 12% to 30%. A two-way, repeated
measures analysis of variance for the total words correct
revealed significant main effects for processing (AFMA > fixed),
F(1, 54) = 15.53, p < .0001; significant main effects for
noise level, F(2, 54) = 10.60, p < .0001; and no significant
Figure 3. Average speech recognition scores across all noise
levels (top) and as a function of noise level for total words correct
( bottom) on the Hearing in Noise Test for adaptive FM advantage
(AFMA) and fixed FM advantage systems. Error bars reflect ±1 SD.
interaction, F(2, 54) = 2.58, p > .05. To compare with the HINT
results, follow-up tests with Bonferonni adjustments for multiple t tests (Sankoh et al., 1997) were performed. As before,
AFMA processing was significantly better than fixed for
the 68- and 73-dBA noise levels, as shown in Table 6. As
expected on the basis of the HINT results, benefit provided
by AFMA processing (as determined by total correct on
the SPIN) significantly increased with noise level from 8%
improvement at 63 dBA to 36% improvement at 73 dBA.
It was also of interest to determine whether the lowpredictability sentences were more sensitive than the highpredictability sentences to differences in the FM processing
schemes. The additional benefits of AFMA processing may
only have been useful when there were no contextual cues
to use. For each noise level, a two-way, repeated measures
Figure 4. Average speech recognition scores for total correct
(top) and for high- and low-predictability sentences (bottom)
across noise levels on the Speech Perception in Noise Test
for AFMA and fixed FM advantage systems. Error bars reflect
±1 SD. A-High = adaptive FM high-predictability sentences;
A-Low = adaptive FM low-predictability sentences; F-High =
fixed FM high-predictability sentences; F-Low = fixed FM
low-predictability sentences.
42 American Journal of Audiology • Vol. 19 • 36–45 • June 2010
Downloaded from aja.asha.org on June 14, 2010
analysis of variance was completed to examine the effects
of context and processing scheme. There was a main effect
for processing (AFMA > fixed) at the two higher noise levels
as expected: 68 dBA, F(1, 36) = 13.42, p < .0001; 73 dBA,
F(1, 36) = 18.84, p < .0001. There were no significant differences between sentence type (high- vs. low-predictability)
at any of the noise levels: 63 dBA, F(1, 36) = 1.77, p > .01;
68 dBA, F(1, 36) = 5.63, p > .01; and 73 dBA, F(1, 36) = 0.97,
p > .01. All interactions were nonsignificant. The AFMA
processing provided significant benefit at the high noise
levels regarding the presence of contextual cues.
To examine how many individuals benefited from the
AFMA processing, 95% confidence intervals were determined for each comparison (AFMA vs. fixed) at each noise
level. The number of participants whose benefit exceeded the
95% confidence interval is provided in Tables 5 and 6. As
expected, the number of participants showing significant
benefit from AFMA processing increased with increasing
noise level. Eight of the 10 individuals made significant
improvements with the HINT sentences via AFMA processing at one of the highest two noise levels. For the SPIN,
the AFMA processing was significantly better for nine of the
10 participants on the low-predictability sentences of the
SPIN at the 73-dBA noise level. For the high-predictability
sentences and total correct, seven of the 10 participants
showed significant benefit with the AFMA processing at the
73-dBA noise level. This shows that the majority of listeners
were receiving significant benefit from the AFMA processing rather than only a few listeners making very large gains.
Subjective Results
The participants selected a preferred processing (AFMA
vs. fixed) at the conclusion of each activity, which is shown in
the right columns of Table 4. For the classroom activities,
100% of the participants chose AFMA as the preferred processing. AFMA was the preferred processing in the aquarium
by all participants at two of the stations. For the remaining
stations, AFMA was also preferred by 80%–90% of the participants. The noise levels at the stations varied between 70
and 90 dBA because of the unpredictable animal sounds as
well as nearby discussions by other visitors. The preferences
may have been affected by nonauditory factors, such as
challenging vocabulary in an unfamiliar context or distracting
movements by nearby animals.
Participant comments regarding the AFMA processing
were that it generally sounded clearer, louder, and easier to
understand. One listener reported, “With AFMA, the voice is
easier to distinguish when in the background noise.” General
comments regarding the fixed processing were that noise
was too loud and that the talker was difficult to hear. Comments
included that the “voice sounded muted” and “it was like the
guide’s voice disappeared when I switched to the B [fixed]
setting.”
Summary and Discussion
The AFMA processing resulted in significant improvements
at the higher noise levels for the majority of the participants.
Despite the inconvenience of the prototype body-worn
processor, the participants were enthusiastic about the new
processing both in clinical and real-world settings. The benefits increased with increasing noise levels for both HINT and
SPIN measures and reached a maximum improvement of
50% at the 73-dBA noise level with the HINT sentences. Differences between the two types of FM processing were not
expected at the low background noise levels because the
adaptive feature is activated only when the noise reaches
57 dB SPL. In quieter environments, the adaptive and fixed
FM systems are operating similarly. AFMA was also the
preferred processing by most of the listeners in a variety of
listening settings. On the basis of these findings, it is likely
that the AFMA processing would provide significant benefits
over fixed FM advantage processing in environments where
the noise exceeds 57 dB SPL. Although these benefits are
expected with the increased FM advantage, it is important that
these listeners confirmed that the adaptive changes were
acceptable.
The participants were given the opportunity to select a
comfortable starting level in quiet. This was necessary because the DAI connections varied in impedance match between the receiver and the hearing aid, a condition for which
the prototype FM receivers used in this study were unable to
compensate. This mismatch in impedance would result in
variations in signal strength. By allowing the listeners to
select a comfortable listening level at the start, these differences were minimized. The fact that there were no significant
differences between processing types at the low noise level
(63 dBA) suggested that there was not inequality between the
initial signal level in the two processing schemes at the low
noise levels. The differences in FM advantage that led to
improvements in performance came into effect at the higher
noise levels (68 and 73 dBA), suggesting that the AFMA
processing had resulted in an increased FM advantage after
sampling the noise and determining that it exceeded the
57-dB SPL criteria.
It is important to note that for the adaptive processing to
work, the transmitter and receiver must both have the adaptive circuitry. Although FM receivers with fixed FM advantages can receive signals from FM transmitters with AFMA
processing, the benefits of the AFMA will not be realized.
Likewise, speech recognition benefits will not be accomplished with FM transmitters with fixed FM advantage when
used with FM receivers with AFMA. This is important to
consider as programs develop plans for upgrading equipment.
It is of interest to compare the performance with AFMA
with that obtained in previous research with FM systems.
Boothroyd (2004) measured the percentage of correct phoneme recognition in 12 adults with mild-to-severe hearing
loss in a word recognition task. Boothroyd found a maximum
performance level when using ear-level FM systems of approximately 80% in 55 dB SPL of noise. With adaptive FM
in the present study, participants perceived 90% of the first
words correct and 85% of the total words correct for the HINT
sentences presented in 54 dBA of noise. With the fixed FM
advantage, the participants recognized 88% and 89% of the
first words and total words, respectively. It is also important
to note that Boothroyd’s results were obtained with FM-only
settings rather than the FM-plus-microphone setting used
in the current study. The better performance in the current
study was expected because nonlinear, multiple-channel,
Thibodeau: Adaptive FM
Downloaded from aja.asha.org on June 14, 2010
43
digital instruments were used compared with the fixed FM
advantage with linear, single-channel, and analog behindthe-ear instruments used in Boothroyd’s study.
Similar benefits with AFMA processing have been reported
by Wolfe et al. (2009) for listeners with Cochlear Corporation
and Advanced Bionics cochlear implants who were also
tested with HINT sentences and the same background classroom noise. Differences in speech recognition of HINT sentences when listening with the adaptive and fixed FM were
only present at higher noise levels of 65, 70, and 75 dBA. As
expected, listeners with cochlear implants had very limited
speech recognition at the highest noise levels, particularly
with the fixed FM system in which they could repeat less than
1% of the words in HINT sentences. However, with the adaptive FM, they repeated 19% of the words. Similarly, for
listeners with hearing aids, speech recognition at the 73-dBA
noise level increased from 5% with fixed FM advantage to
55% with AFMA processing.
The greater increase with adaptive FM processing for
hearing aid users (50%) compared with the increase obtained
by cochlear implant users (18%) suggests that the hearing
aid users were better able to use the increased acoustic cues
available through the adaptive FM processing. In fact, for
the highest noise level in the present study, 80 dBA, none of
the listeners were able to repeat any of the HINT sentences
with the fixed FM processing, yet seven of them were able to
repeat on average 22.7% of the words correct with the adaptive FM processing. The greatest improvement was made
by a student (Participant 7 in Table 3) who could not repeat
any words with fixed FM processing and who was able
to repeat 64% correct with adaptive FM processing at the
80-dBA noise level. One adult (Participant 8 in Table 3) also
showed remarkable performance for fixed processing with
the SPIN at the highest level. For this highest noise level
(73 dBA), none of the listeners could repeat any of the highpredictability sentences with the fixed processing, yet she
was able to repeat them all. However, with AFMA processing at that noise level, she only repeated 60% correct. She
attributes her excellent speech recognition in noise to the
fact that she had many years of “straining to hear” because
she did not get hearing aids until she was in college.
Listeners with hearing aids and cochlear implants showed
significant improvements with the adaptive FM processing.
This FM processing will be of considerable benefit in noisy
classrooms where average levels are as high as 68 dBA
(Crandell & Smaldino, 2000). With the use of this new FM
processing, listeners with hearing loss may also experience
benefits in environments where communication was not
even possible, such as sporting events, restaurants, shopping
malls, or social gatherings. In fact, one participant (Participant 8) reported being able to hear directions for the first time
in an aerobics class when the instructor wore the transmitter rather than simply following the exercise moves on the
basis of visual cues. Reports of average noise levels in public
venues are limited; however, when available, the data suggest that AFMA processing in an FM system could have a
significant beneficial impact on communication. For example, general communication while standing on a street could
be improved with a transmitter with AFMA processing on the
talker and a compatible FM receiver on the listener. Seto,
Holt, Rivard, and Bhatia (2007) reported that the traffic noise
level in San Francisco averaged over day and night was 68 dB,
which exceeds the 57-dBA criteria to activate the AFMA
processing. This level varies with residential or commercial
areas as well as the type of traffic, such as cars or trucks.
Another setting where use of an FM system with AFMA
processing would be useful is in a subway. Noise levels in
subways were recorded as 90–155 dBA, which could certainly
interfere with critical information exchange at a ticket counter
(Northwestern Health Sciences University, 2001). Although
more social communication, use of AFMA FM processing
would be helpful at football games where noise is reportedly
117 dBA (Northwestern Health Sciences University, 2001)
or in a hair salon where even just one hair dryer was measured
at 90 dBA (National Institute for Occupational Safety and
Health, 2008). Finally, a daily experience for many persons
with hearing loss is communicating in a car where internal
noise could be 71 dBA (Noise Levels, 2000) before any additional factors, such as a radio or active children, are added.
According to the results of the current study, use of AFMA
FM processing in any of these settings could result in significantly better speech recognition. Further research will be
needed with the final products to ensure that the benefits of
AFMA processing observed in the current study at the higher
noise levels are maintained when the FM receivers are coupled
to newer hearing aids as well as with other types of cochlear
implants.
Acknowledgments
The assistance by Martin Luetzen while working at Phonak is
gratefully acknowledged. Data collection and analysis were facilitated
by University of Texas at Dallas doctoral students Lindsay Bondurant,
Jessica Sullivan, Gregory Nelson, Sarah Tillman, and Ivonne
Perez-Cervantes. The cooperation of staff at the Dallas World
Aquarium is also gratefully appreciated.
References
American Academy of Audiology. (2008). AAA clinical practice guidelines: Remote microphone hearing assistance technologies for children and youth birth–21 years. Retrieved from
www.audiology.org/resources/documentlibrary/documents/
hatguideline.pdf.
American Speech-Language-Hearing Association. (2002).
Guidelines for fitting and monitoring of FM systems. Available
from www.asha.org/policy.
Bilger, S., Nuetzel, J. M., Rabinowitz, W. M., & Rzeczkowski, C.
(1984). Standardization of a test of speech perception in noise.
Journal of Speech and Hearing Research, 27, 32–48.
Boothroyd, A. (2004). Hearing aid accessories for adults: The
remote FM microphone. Ear and Hearing, 25, 22–33.
Crandell, C., & Smaldino, J. (2000). Classroom acoustics for
children with normal hearing and with hearing impairment.
Language, Speech, and Hearing Services in Schools, 31, 362–370.
Dunn, L., & Dunn, L. (1981). Peabody Picture Vocabulary Test—
Revised. Circle Pines, MN: AGS.
Hawkins, D. B. (1984). Comparisons of speech recognition in
noise by mildly-to-moderately hearing-impaired children using
hearing aids and FM systems. Journal of Speech and Hearing
Disorders, 49, 409–418.
44 American Journal of Audiology • Vol. 19 • 36–45 • June 2010
Downloaded from aja.asha.org on June 14, 2010
Lewis, D., & Eiten, L. (2004). Assessment of advanced hearing
instrument and FM technology. In D. A. Fabry & C. DeConde
Johnson (Eds.), ACCESS: Achieving Clear Communication
Employing Sound Solutions—2003. Proceedings of the First
International FM Conference (pp. 167–174). Warrenville, IL:
Phonak AG.
Lewis, M. S., Crandell, C. C., Valente, M., & Enrietto Horn, J.
(2004). Speech perception in noise: Directional microphones
versus frequency modulation (FM) systems. Journal of the
American Academy of Audiology, 6, 426–439.
National Institute for Occupational Safety and Health. (2008).
Noise and hearing loss prevention. Retrieved from www.cdc.
gov/niosh/topics/noise/abouthlp/noisemeter_html /hp80.html.
Nilsson, M., Soli, S., & Sullivan, J. (1994). Development of the
Hearing in Noise Test for the measurement of speech reception
threshold in quiet and in noise. The Journal of the Acoustical
Society of America, 95, 1085–1099.
Noise levels for motor vehicles. (2000). Retrieved from www.
xs4all.nl/~rigolett/ENGELS/typetest/carfr.htm.
Northwestern Health Sciences University. (2001). Noise in the
workplace—What decibel levels are safe? Retrieved from www.
nwhealth.edu/healthyU/stayHealthy/ear3.html.
Sankoh, A., Huque, M., & Dubey, S. (1997). Some comments
on frequently used multiple endpoint adjustments methods in
clinical trials. Statistics in Medicine, 16, 2529–2542.
Schafer, E., & Thibodeau, L. (2003). Speech recognition performance of children using cochlear implants and FM systems.
Journal of Educational Audiology, 11, 15–26.
Schafer, E., & Thibodeau, L. (2004). Speech recognition abilities
of adults using cochlear implants with FM systems. Journal
of the American Academy of Audiology, 15, 678–691.
Schafer, E., & Thibodeau, L. (2006). Speech recognition in noise
in children with cochlear implants while listening in bilateral,
bimodal, and FM-system arrangements. American Journal of
Audiology, 15, 114–126.
Schafer, E., Thibodeau, L., Whalen, H., & Overson, G. (2007).
Electroacoustic evaluation of frequency-modulated receivers
interfaced with personal hearing aids. Language, Speech, and
Hearing Services in Schools, 38, 1–12.
Seto, E., Holt, A., Rivard, T., & Bhatia, R. (2007). Spatial distribution of traffic induced noise exposures in a U.S. city: An
analytic tool for assessing the health impacts of urban planning
decisions. International Journal of Health Geographics, 6, 24.
Thibodeau, L. (1990). Electroacoustic performance of twenty
direct-input hearing aids with two FM amplification systems.
Language, Speech, and Hearing Services in Schools, 21, 49–56.
Thibodeau, L. (2004). Terminology and standardization. In D. A.
Fabry & C. DeConde Johnson (Eds.), ACCESS: Achieving Clear
Communication Employing Sound Solutions—2003. Proceedings of the First International FM Conference (pp. 75–87).
Warrenville, IL: Phonak AG.
Thibodeau, L., McCaffrey, H., & Abrahamson, J. (1988). Effects of coupling hearing aids to FM systems via neckloops.
Journal of the Academy of Rehabilitative Audiology, 21, 49–56.
Thibodeau, L., & Saucedo, K. (1990). Consistency of electroacoustic characteristics across components of FM systems.
Journal of Speech and Hearing Research, 34, 628–635.
Wolfe, J., Schafer, E., Heldner, B., Mulder, H., Ward, E., &
Vincent, B. (2009). Evaluation of speech recognition in noise
with cochlear implants and dynamic FM. Journal of the
American Academy of Audiology, 20, 409–421.
Received June 4, 2009
Revision received December 23, 2009
Accepted March 8, 2010
DOI: 10.1044/1059-0889(2010/09-0014)
Contact author: Linda Thibodeau, Advanced Hearing Research
Center, University of Texas at Dallas, 1966 Inwood Road,
Dallas, TX 75080. E-mail: [email protected].
Thibodeau: Adaptive FM
Downloaded from aja.asha.org on June 14, 2010
45
Benefits of Adaptive FM Systems on Speech Recognition in Noise for Listeners Who Use
Hearing Aids
Linda Thibodeau
Am J Audiol 2010;19;36-45; originally published online Mar 10, 2010;
DOI: 10.1044/1059-0889(2010/09-0014)
The references for this article include 4 HighWire-hosted articles which you can access for free
at: http://aja.asha.org/cgi/content/full/19/1/36#BIBL
This information is current as of June 14, 2010
This article, along with updated information and services, is
located on the World Wide Web at:
http://aja.asha.org/cgi/content/full/19/1/36
Downloaded from aja.asha.org on June 14, 2010