Download The measurement of auditory profiles

Tinnitus
Hearing
Aids
Speech in
Noise
Model
Data
Hearing
Dummy
Project
Data
Collection
Essex Hearing Dummy Project
Auditory profiles
(January 2012)
Prof. Ray Meddis
Dr. Wendy Lecluyse
Dr. Christine Tan
Dr. Nick Clark
Dr. Tim Juergens
1
Table of contents
Introduction ............................................................................................................................................ 3
The measurement of auditory profiles ................................................................................................... 4
IFMC ‘depth’ measure............................................................................................................................. 5
TMC slope measure ................................................................................................................................ 5
Database summary ............................................................................................................................... 12
Average profiles .................................................................................................................................... 13
Statistical analyses ................................................................................................................................ 16
Comparing good and impaired hearing statistics ................................................................................. 17
Raw profiles (Normal) ........................................................................................................................... 19
Raw profiles (impaired)......................................................................................................................... 32
Double profiles ...................................................................................................................................... 78
File format and programs ..................................................................................................................... 86
2
Introduction
The Hearing Laboratory at Essex has accumulated a useful number of auditory profiles for a range of
people with good hearing or sensorineural hearing impairment. These profiles were collected using
the multiThreshold software. This document describes how to access these profiles. The data were
collected by Dr Wendy Lecluyse and Dr Christine Tan as part of their doctoral studies. Dr. Tan
transcribed the data from the individual participant records to a more standard format using
MATLAB .m files.
The profiles were collected from individuals with a sensorineural hearing loss (average age 59 years)
as well as some young adults with good hearing (average age 32 years). Both groups consist of
unpaid volunteers. Altogether, there are 77 participants; 23 with good hearing and 54 with impaired
hearing. The average ages (with standard deviations) are, respectively, 32 (10) and 59 (11) years. The
male/female ratio is approximately 3:2 in both cases.
The profiles and participant information are held in a profiles folder in the MAP1_14 software
package. A readable representation of all the profiles, each presented as a separate chart can be
found in a WORD document, Profiles summary.doc.This document also contains some statistical
analysis. For many, this will be all that is required to become familiar with the dataset. However, for
those who wish to investigate the profiles in greater detail an appendix is supplied, describing the
programs used to organise and analyse the data.
3
The measurement of auditory profiles
An auditory profile summarises the data generated using three tests measuring absolute threshold,
frequency-selectivity and compression. A detailed account of the procedures is given below. The
data for a single individual is combined to generate a graphical representation called an auditory
profile. This section offers a detailed guide to these profiles
masker dB SPL
0.25
0.5
dB SPL
2
4
6
50
0
mean
51
77
0
100
gap (ms)
100
1
100
15
82
40
80
60
mean
NH25_R /
21
36
65
47
42 34
32
50
18
0
.25
.5
1
2
frequency (Hz)
4
8
This profile was collected from one of the authors with good hearing. All levels are dB SPL.
The top row shows the temporal masking curve (TMC) measured at a number of different probe
frequencies. The probe frequency is indicated at the top of each panel. The slope of the TMC is
supposed to be an indication of compression (steeper slopes, more compression). The original data
points are the unfilled circles. The line is the best-fit straight line. The slope of the line is printed at
the bottom of the panel in terms of the dB increase in the masker level per 100-ms increase in the
gap between the masker and probe. The average slope across all frequencies is given to the right.
The bottom panel shows the V-shaped frequency-selectivity curves for probes of different
frequencies. These are iso-forward-masking contours (IFMCs). The probe frequency (fp) is indicated
by the unfilled circles along the function. However the points along the function are the masker
levels at different frequencies. The ‘depth’ of the V is posted at the top of the panel above the
corresponding function. It represents the difference between the level of the masker at fp and the
average of masker levels at 0.7*fp and 1.3*fp. The mean depth across all frequencies is shown to the
right.
The lines at the bottom of the panel show the absolute thresholds for a 16-ms probe and (below it)
the absolute threshold for a 250-ms probe at frequencies 250, 500, 1000, 2000, 4000 and 8000. The
mean absolute threshold is shown to the right.
4
IFMC ‘depth’ measure
The depth measure is defined as ‘the difference between the level of the masker at fp and the
average of masker levels at 0.7*fp and 1.3*fp’ and is illustrated below. Clearly, this statistic will be
large when the arms of the V-shape are steep and when the V is symmetric. Any deviation from this
pattern will reduce the depth estimate and even render it negative.
profile_NH80_R
profile_IH4_L
100
masker level (dB SPL)
masker level (dB SPL)
100
80
60
depth= 25
40
20
0
0.7
1
80
depth= -2
60
40
20
0
1.3
0.7
masker frequency ratio
1
1.3
masker frequency ratio
TMC slope measure
The slope is simply the slope of the least-squares best-fit straight line. In almost all cases, this is a
good fit.
profile_NH80_R
profile_IH4_L
100
masker level (dB SPL)
masker level (dB SPL)
100
80
60
40
20
slope= 26 dB/100 ms
0
0
20
40
60
80
masker-frequency gap (ms)
100
80
60
40
20
slope= 12 dB/100 ms
0
0
20
40
60
80
100
masker-frequency gap (ms)
Occasionally the line has a distinct bend which may indicate a real change in the underlying function.
This may indicate a change from a linear to a compressed region. In these rare cases the slope is an
underestimate of the compressed portion of the line.
5
profile_NH84_L
masker level (dB SPL)
100
80
60
40
20
slope= 58 dB/100 ms
0
0
20
40
60
80
masker-frequency gap (ms)
6
100
Equipment
Absolute threshold, frequency selectivity and compression tests were carried out in a sound-proof
booth. Stimuli were presented through circumaural headphones (Sennheiser HD600) linked directly
to a computer sound card (Audiophile 2496, 24-bit, 96000-Hz sampling rate). The procedures were
automated using a computer program written in the MATLAB computer language. This program is
available from the authors on request. Participants were equipped with a small console with 4
buttons. A computer monitor in front of the participant showed a graphical user interface display of
the button console. While the stimulus was presented the button symbols on the display
disappeared. Immediately after stimulus presentation the buttons reappeared on the screen
signalling that a response was required.
Procedure and stimuli
All stimuli used pure tones, ramped with raised cosine onset and offset times of 4 ms.
Schematic representations of the stimuli and the resulting thresholds are shown in Fig. 1. All stimuli
(tone alone or masker-probe tone combinations) were preceded by cue stimuli that were identical to
the test stimuli in all respects except for a single difference arranged so that the target tone was
always more audible in the cue stimulus. For absolute threshold measurements, the cue tone was
always 10 dB more intense than the test tone (Fig 1A). For frequency selectivity and compression
measures, the cue masker was always 10 dB less intense than the test masker (Fig 1B and C). The
cue-test stimulus interval was 500 ms.
Absolute thresholds were measured using 250-ms pure tones (Fig. 1A) at frequencies (ft) 250,
500, 1000, 2000, 4000 and 8000 Hz.
Frequency selectivity was assessed using forward masking. The patient’s task was to report
the presence or absence of a probe tone following a masking tone. Each measurement was a
masking threshold, the masker level required to just mask the probe tone. Masking thresholds were
measured at 7 different masker frequencies (fm) specified relative to the probe frequency (0.5, 0.7,
0.9, 1, 1.1, 1.3, and 1.6 * ft) to generate an Iso-Forward Masking Contour (IFMC). IFMCs were
determined for a range of probe frequencies (ft), 500, 1000, 2000, 4000 and 6000 Hz. Masker and
probe durations (Dm, Dt) were 108 ms and 16 ms respectively. For each probe frequency the absolute
threshold for the 16-ms probe tone needed to be determined. Subsequently this threshold was used
to fix the level of the probe in the forward masking task. Here the probe tone was always presented
at 10 dB above its threshold. The gap duration between masker and probe was set at 10 ms. A
schematic representation of the stimulus and an example of a 2000-Hz IFMC are shown in Fig. 1B.
Compression was also assessed forward masking. Similar to the IFMC, the patient’s task was
to report the presence or absence of a probe tone following a masking tone. Each measurement
comprised of a masking threshold. Masking thresholds were measured at 5 different gap durations
(20, 40, 50, 60 and 80 ms) to generate a Temporal Masking Curve (TMC). Gap duration was defined
as the duration of the silence period between the masker offset and the probe onset. TMCs were
obtained for probe frequencies (ft) of 500, 1000, 2000, 4000 and 6000 Hz. Masker frequency (fm) was
set equal to the probe frequency. Masker and probe durations were 108 ms and 16 ms respectively.
7
For all conditions, the probe level was fixed at 10 dB above its own threshold. A schematic
representation of the stimulus and an example of a TMC are shown in Fig. 1C.
The main difference in the protocol described above with the longer, initial protocol (phase
I) is that each data point was the result of a single measurement. In phase I, data points were
generally the average of 3 measurements. Some smaller, less important changes were adjustments
of the tone duration for the absolute thresholds (250 ms instead of 500 ms) and adjustment of the
probe duration in the forward masking tasks (16 ms instead of 8 ms). The latter adjustment
increased the ease of the forward masking task. Finally, the Phase I TMC consisted of 9 data points
(10-90 ms gaps in 10-ms intervals) instead of the 5 data points used in later tests. Rather than omit
these additional points in the interests of consistency, these data points were retained in the profiles
to be presented.
Pure-tone Audiometry was conducted using standard protocols (BSA 2004). Audiometric
thresholds were measured at 250, 500, 1000, 2000, 4000 and 8000 Hz.
Threshold estimation procedure
The basic procedure for measuring absolute and masking thresholds was based on an adaptive yesno paradigm (Bekesy 1947, Dixon & Mood 1948, Carhart & Jerger 1959, Cornsweet 1962, Leek et al.
2000). This single-interval up/down procedure (SIUD), was based on modifications previously
proposed and evaluated by Lecluyse and Meddis (2009).
A test stimulus was presented to the participant whose response was interpreted as ‘yes’ or
‘no’, i.e. whether or not the test tone was heard. A level adjustment was applied from trial to trial
using a one-down, one-up adaptive procedure. The direction of the adjustment depended on the
task participants were asked to perform. When measuring absolute threshold, the stimulus level was
increased by a fixed amount if the response was a ‘no’. If the response was a ‘yes’, the level was
decreased by the same amount. In the forward masking tasks the masker level was increased when
the test probe was detected and decreased when no test probe was heard. By presenting a cue
stimulus before the test stimulus, this task now became a counting task, with the listener indicating
‘how many’ tones were heard. A ‘2’ response indicated that both the cue and the test stimulus had
been heard, equivalent to a ‘yes’ response. A ‘1’ response meant only the easier-to-hear cue was
perceived and corresponded to a ‘no’ response.
The threshold run started with an initial phase where the stimulus level was set at a level
generating a guaranteed ‘yes’-response and was adjusted using a 10-dB step size until the first ‘no’response. After the first ‘no’ response, the stimulus level was set to the mid-point between the
previous two levels and a smaller, 2-dB, step size was used thereafter. The run then continued for 10
trials counting from the trial immediately before the first ‘no’ response.
8
Cue + test stimulus
Data (schematic)
100
Test tone
Threshold (dB SPL)
Cue tone
10 dB
A.
ft
ft
//
Absolute
threshold
time
500
ms
80
60
40
20
0
250 ms
Cue masker Cue probe
Test masker Test probe
10 dB
//
ft
B. IFMC
variable f m
ft
time
500
ms
108 ms
16 ms
108 ms
10 ms
16 ms
Test masker Test probe
10 dB
//
ft
C. TMC
16 ms
variable gap
9
fm
ft
500
ms
108 ms
0
1000 2000 3000 4000
Masker frequency (Hz)
0
0.02 0.04 0.06 0.08 0.1
Gap duration (s)
80
60
40
20
10 ms
Cue masker Cue probe
fm
1000
10000
Tone frequency (Hz)
0
108 ms
16 ms
variable gap
time
100
Masker threshold (dB SPL)
variable f m
100
100
Masker threshold (dB SPL)
250 ms
80
60
40
20
0
The start level of the stimulus was different in each run and randomly located in a range ±5
dB relative to the nominal start value. For the forward masking measures (IFMC and TMC) the start
value of the masker was set at a low level such that only the probe tones were heard. This ensured
that the listener was reminded what the probe tone sounded like and where in time it would occur.
The threshold was estimated at the end of the run. All stimulus levels from the trial before
the first reversal onwards were included in the estimate of the threshold. Earlier trials were
discarded. The threshold was estimated by fitting a psychometric function of the form p(L)=
1/(1+exp(-k(L-θ))), where p is a binary vector of the yes/no responses, L is a vector of the levels (dB
SPL) associated with the response vector, k is a slope parameter and θ (dB SPL) is the threshold to be
estimated. The threshold, θ, is the level of the stimulus at which the proportion of yes-responses is
expected to be 0.5. The psychometric function was fitted to the responses using a least-squares,
best-fit procedure, with θ and k as free parameters.
Catch trials. One in five trials were ‘catch trials’ where the cue tone was retained but the test
tone was omitted. In this case a ‘0’ or a ‘1’ response was treated as correct but a ‘2’ response was
incorrect (false-positive). This was taken to indicate that the listener was not attending or was using
an inappropriate strategy. If the participant produces a false-positive, the run was stopped and
restarted; possibly after resting the participant and giving further instructions. Participants are
encouraged not to guess but to report hearing a (possibly very faint) test tone only when they are
confident that they have heard it. The restart process following the false-positives acts as an
additional incentive for patients to make only confident judgments. A catch trial was always
presented on the 2nd trial in a run to remind the listener of what a ‘no-stimulus’ sounds like. Catch
trials are not included in the trial count.
10
Instruction and training
The instructions given to the listener need to be carefully worded and can be found in
appendix A. Experienced experimenters have been able to generate reliable data very quickly (see
below). However an element of skill might be involved in instructing the participants. Therefore a
clear set of written instructions was assembled and this was found to enhance reliability.
Pilot experiments (not shown) investigating the training effects in the SIUD-procedure
showed that a single run in the absolute threshold task was sufficient to provide stable threshold
estimates. Similarly, stable masking thresholds were obtained after only 4 threshold runs. Therefore
no formal training session took place. Participants received instructions on the absolute threshold
task and had a trial run which familiarized them with the testing procedure. If necessary, this was
repeated until the participant felt confident about the task at hand. Data collection always started
with the absolute threshold measurements. A similar process was followed for the forward masking
task. The forward masking task was always introduced in the context of the IFMC measures. Again,
the participant was instructed on the task and subsequently performed a trial run. A number of
points on the IFMC were found to be more suitable as training conditions such as the condition with
masker frequency = 0.5 * probe frequency. The substantial ‘distance’ in frequency between the
masker and the probe tone improves the contrast between the 2 stimuli and facilitates the decision
concerning the presence or absence of the probe tone in the presence of the masker. Once the
participant was confident about the forward-masking task, data collection continued. TMC
measurements were always the last section conducted in the protocol.
11
Database summary
Participant summary
number of participants = 103
number of impaired /normal hearing used= 65 / 25
number of males /females = 56 / 34
number of impaired males /females = 41 / 24
number of normal males/females = 15 / 10
average age impaired/ normal = 59.2 (11.0) / 32.3 (9.8 )
impaired with tinnitus yes/no/not determined = 44 / 19 / 2
normal with tinnitus yes/no/not determined = 3 / 22 / 0
117 ears measured
89 impaired ears measured
28 good ears measured
impaired right ear/ left ear = 47 / 42
normal right ear/ left ear = 19 / 9
12
Average profiles
Profiles were averaged across listeners as follows
The average absolute thresholds were simple averages across subjects.
The average TMCs were simple averages across subjects.
The average IFMC was obtained by first expressing all levels relative to the masker level at the probe
frequency, averaging and then adding back the average threshold at the probe frequency.
13
Normal
250
500
1000
2000
4000
0.25 0.5
6000
masker dB SPL
masker dB SPL
100
80
60
40
20
62
60
59
52
66
64
meanNHprofileTest: mean IFMC SD is 12.4 dB
100
30
11
30
19
31
28
30
26
50
4
6
50
mean
62 60 59 52 66 64
0 (17) (24) (24) (26) (26) (26)
0 100
gap (ms)
100
26
23
dB SPL
IFMC N= 20
depth=4
.
2
100
0
0
100
gap (ms)
1
60
mean
~im paired
4
11 19 28 26 23
(20) (30) (30) (31) (30)(26)
19
50
19
0
0
2
10
3
.25 .5
1
2
4
probe frequency (Hz)
4
10
probe frequency (Hz)
10
8
Impaired
250
500
1000
2000
4000
0.25 0.5
6000
masker dB SPL
masker dB SPL
100
80
60
40
20
11
10
13
20
18
10
meanIHprofileTest: mean IFMC SD is 17.1 dB
100
73
7
81
11
82
14
68
10
50
4
6
50
mean
11 10 13 20 18 10
0 (24) (52) (60) (61) (55) (39)
0 100
gap (ms)
100
48
12
dB SPL
IFMC N= 47
depth=3
.
2
100
0
0
100
gap (ms)
1
14
mean
im paired
3
6
11 14 10 11
(47) (73) (81) (82) (69)(48)
50
47
0
0
2
10
14
3
10
probe frequency (Hz)
4
10
9
.25 .5
1
2
4
probe frequency (Hz)
8
With tinnitus
500
1000
2000
4000
6000
250
100
100
80
80
masker dB SPL
masker dB SPL
250
without
60
40
20
17
17
26
28
20
2000
4000
6000
5
13
17
-7
40
20
4
6
0
0
100
gap (ms)
meanTinnitusProfile: mean IFMC SD is 8.7 dB
0
100
gap (ms)
meanNoTinnitusProfile: mean IFMC SD is 4.9 dB
.
100
.
100
IFMC N= 28
depth=4
42
9
49
14
51
17
44
13
31
13
IFMC N= 18
depth=1
50
29
4
30
7
29
8
23
6
16
10
50
0
2
10
3
0.5
1
2
0
2
10
4
10
probe frequency (Hz)
0.25
10
4
0.25
17
17
26
27
20
25
0
1
2
4
6
50
0
100
12
42
0.5
mean
4
6
5
13
17
-7
0 100
gap (ms)m eanNoTinnitusProfile
mean
50
1
4
7
8
8
6
mean
6 10
6
50
48
0
.25 .5
1
2
4
probe frequency (Hz)
15
22
4
10
100
masker dB SPL
mean
dB SPL
masker dB SPL
50
0
3
10
probe frequency (Hz)
6
100
0 100
gap (ms) m eanTinnitusProfile
100
4
8
14 18 12 13
dB SPL
1000
60
25
0
500
.25 .5
1
2
4
probe frequency (Hz)
8
Statistical analyses
16
Comparing good and impaired hearing statistics
This figure shows the average thresholds, slopes and frequency-selectivity depth for two sets of
profiles (good hearing and impaired hearing) at a range of different frequencies.
After removing NH with low-level linearity: 5,6,8,16,22,23
The histograms in the figures that follow show the distribution of threshold, slope and depth
measures for the two groups separately.
The scatter plots show the correlations among the three measurements, again computed
separately for the two groups and for six probe frequencies
17
18
Raw profiles
(Normal)
All profiles are shown.
When only one ear was tested the other ear is left blank.
19
20
21
22
23
24
25
26
27
analyse
accumulated data for this group & display average25participants
SS =
IFMCsampleSize: [20 30 30 31 30 26 11]
absThrSampleSize: [41 41 41 41 41 23 40]
TMCsampleSize: [17 24 24 26 26 26 5]
meanLongTones: [14.5715 8.9344 5.1632 5.6451 4.2698 13.4435 20.0733]
meanShortTones: [1x7 double]
stdevLongTones: [6.1229 5.3700 6.6323 8.0837 6.3709 5.9082 10.5285]
stdevShortTones: [4.3140 4.6605 4.2868 6.2639 5.5783 5.0014 8.7179]
stdevIFMCs: [7x7 double]
+ stdevTMCs: [7x9 double]mean short vs long= 250
8.74000
8.36000
5.88000
28
10.4 500
10.31000
10.42000
29
30
hist
Compare left and right ears
mean abs threshold difference= 0.46571 logical requ
31
Raw profiles
(impaired)
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
analyse accumulated data for this group & display average62participants
SS =
IFMCsampleSize: [47 73 81 82 69 48 30]
absThrSampleSize: [119 119 118 114 108 73 85]
TMCsampleSize: [24 52 60 61 55 39 20]
meanLongTones: [1x7 double]
meanShortTones: [1x7 double]
stdevLongTones: [1x7 double]
stdevShortTones: [1x7 double]
stdevIFMCs: [7x7 double]
stdevTMCs: [7x9 double]
mean short vs long= 250
4.36000
4.28000
54
9.0 500
8.81000
6.52000
5.94000
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
hist
Compare left and right ears mean abs threshold difference= 0.46571 logical requirement= impaired
77
Double profiles
(normal and impaired)
All profiles are shown for cases where both left and right ear were studied in depth.
The purpose is to show similarities between the two ears and offer evidence of the reliability of the
testing process.
78
Left ear
79
right ear
80
81
82
83
84
Excluding unilateral impairments
TMC slope
IFMC depth
60
40
35
50
30
right ear
right ear
40
30
20
10
0
85
r= 0.94
N= 17
0
20
40
left ear
60
25
20
15
10
r= 0.94
5
N= 17
0
0
20
left ear
40
File format and
programs
The files
The complete collection of individual profiles (one per ear per person) is held in a subfolder called
allParticipants. Each profile is held in a separate .m file so that each participant has two files (left and
right ear). If only one ear was studied, the other ear file exists but contains no data. When the .m
file is run it creates and returns a structure containing all of the data.
The name of each data file consists of the word ‘profile_’, followed by the user’s code (e.g. ‘NH03’)
and then and ‘_R’ or ‘_L’ according to which ear was tested; for example, profile_NH96_L is the left
ear profile for participant #96 with normal hearing. Or profile_IH10_R for the left ear of participant
#10 with impaired hearing. The participant number is unique.
These files are ‘raw data’ but even these are digests of more extended investigations. The data in the
.m files are an attempt to introduce uniformity across participants in the presentation of available
data. The original data is stored in Excel spreadsheets along with all of the testing details. These
individual workbooks are available on request.
Participant details are held in an Excel file. The details include participant number, age, gender, year
of testing and whether or not tinnitus is present. The Excel spreadsheet is called ParticipantList.xls.
These data (both the .m profiles and the participantDetails.xls) have been consolidated into a single
data file, called participantCompendium.mat. This compendium file is described later. The
management of individual profile.m files is described first.
Participants are normally tested for three things; absolute thresholds, temporal masking curves
(TMCs) and iso-forward masking contours (IFMCs). These are intended to be measures of sensitivity,
compression and frequency selectivity, respectively and are described more fully in the
multiThreshold Quick reference and Users manual. Detailed testing protocols are described in
Profiles summary.doc.
plotProfile
A single file can be plotted using the plotProfile.m program found in the profiles folder. For example,
the call
plotProfile('allParticipants', 'profile_NH96_L')
plotProfile('allParticipants', 'profile_IH10_L')
86
will generate the two following images
1
2
4
6
0.25
50
0
mean
24
21
53
26
53
47
37
dB SPL
0 100
gap (ms)
profile_NH96_L
100
3
3
25 37 17 29
17
0
8
4
6
mean
74
83
73
35
-1
53
mean
IH10_L /
5
53
50
0
.25 .5
1
2
4
probe frequency (Hz)
2
100
20
0
1
50
0 100
gap (ms)
mean
50
0.5
100
masker dB SPL
100
dB SPL
masker dB SPL
0.25 0.5
2
6
9
6
.25
.5
1
2
frequency (Hz)
1
4
8
This function call has the following format:
plotProfile (<folder name>, <file name>, <comparison file name>, <figure number>)
The last two arguments are optional. The <comparison file name> refers to a second file in the same
folder that is plotted as background (using faint, dashed lines). In the following example, a profile for
someone with a high frequency loss is compared with a young person with good hearing. The final
argument, <figureNumber>, is mainly for use by calling programs.
plotProfile('allParticipants', 'profile_IH10_L', 'profile_NH96_L')
masker dB SPL
0.25 0.5
1
2
4
6
100
50
mean
74 83 73 35 -1
0 24 21 53 26 53 47
0 100
gap (ms) profile_IH10_L / NH96_L
53
mean
dB SPL
100
5
53
50
0
2
3
6
3
9
25
6
37
1
17 2922
.25 .5
1
2
4
probe frequency (Hz)
8
The visual profiles above show the results of attempts to quantify the shapes in the TMC and IFMC
functions. The TMC measure is the slope of the best-fit straight line to the TMC data (in the top
panels) expressed as dB/100ms. This is printed at the bottom of the upper panels. When two rows of
figures are given, the upper row is the first (foreground) file. It can be seen that the good hearing
profile has steeper slopes than the impaired profile but only at higher frequencies.
87
Our crude measure of frequency selectivity is quantified by the ‘depth’ of the V-shapes in the lower
panel. This is computed as the average of the masker level at 0.7 and 1.3 times the probe frequency
minus the masker level at the probe frequency (see diagram below). This simple metric gives large
values when the IFMC is narrow and symmetric. These metrics are meant to be interpreted by
comparison with values obtained from participants with good hearing. In the example above the
good hearing profile has greater depth from 1000 Hz upwards.
The numbers to the right of the plots are the averages across frequency for the first-named profile.
The lowest figure to the right of the lower panel is the average absolute threshold for a long, 250-ms
tone.
88
plotAllFiles
plotAllFiles.m is another program in the profiles folder. This will sequentially and rapidly plot all
profiles of a particular type followed by summary statistics. For example,
plotAllFiles (‘impaired’)
will process and plot all profiles classified as ‘impaired’. At the end of the run an average profile will
be computed. Summary statistics are then computed across all participants for whom relevant data
is available.
Each plot appears in the same figure and they change quickly. The speed of the display can be
controlled by adding a pause duration argument (in seconds),
plotAllFiles('~impaired', 1)
This will slow down the display considerably. This particular instruction will display only files with
normal hearing. This is because impaired is a logical variable so ~impaired refers to all files that are
not impaired. The first argument is a string containing a logical condition evaluated by the program
when selecting from the data in the participantCompendium. The condition possibilities are as
follows:
number
impaired
initials
male
tinnitus
birthYear
startTest
age
code
double (unique subject identifier)
logical
string
logical
logical
double
double (year of testing)
double (age when tested)
string (eg ‘NH03’)
Examples
plotAllFiles('impaired & ~tinnitus', .1)
will plot all impaired files where tinnitus was known not to be present
plotAllFiles('impaired & age>65', .1)
will plot all impaired profiles for those over 65 years of age.
Statistical summaries.
At the end of a run, an average profile is computed and plotted. For example,
plotAllFiles('impaired & age>65',.1)
will generate the following average profile
89
masker dB SPL
0.25 0.5
1
2
4
6
100
50
mean
-3
15 14 19 17 13
0 (8)
12
(23) (25) (24) (16) (9)
dB SPL
0 100
gap (ms) im paired & age>65
100
mean
8
52
50
0
1
6
10
10
10 11
(20) (30) (31) (30) (20)(12)
.25 .5
1
2
4
probe frequency (Hz)
8
The figures in brackets are the number of ears that contributed to the average. This is important
because it is not always possible to measure some values.
Two analysis figures are offered. The first shows the distribution of scores on three variables;
absolute threshold, TMC slope and IFMC depth. These are subdivided according to the frequency
tested. At the top of each subplot, summary statistics are shown; mean, standard deviation and
sample size. Note that the analysis applies only to the profiles that were selected (in this case
participants with impaired hearing aged over 65 years)
BF (Hz)
20
250
m 34 sd16 n=43
10
500
10
2000
5
4000
10
6000
5
0
2
0
m 11 sd=19 n=17
0
m 20 sd18 n=10
1
0 20 40 60 80 100
abs threshold
10
5
0
m 57 sd19 n=22
m 11 sd=13 n=28
0
m 22 sd24 n=16
2
0
10
4
10
5
0
m 55 sd17 n=38
m 11 sd=13 n=30
0
m 23 sd15 n=24
5
0
20
10
20
10
0
m 40 sd22 n=42
m 6 sd=6 n=30
0
m 22 sd19 n=25
5
0
10
10
40
20
0
m 36 sd21 n=43
m 1 sd=4 n=19
0
m 24 sd21 n=23
5
0
1000
10
20
10
0
m 33 sd16 n=43
10
20
m 17 sd25 n=8
2
0
20
4
10
m 13 sd=11 n=10
5
0 20 40 60 80 100
TMC slopes
0
0 20 40 60 80 100
IFMC depth
The second analysis shows scatter plots and correlations for each combination of these three
measures. Again the analysis is performed separately for each probe frequency.
90
100
50
0
IFMC depth
0
50
100
abs threshold(r= -0.35 N=25)
100
50
0
100
50
0
0
50
0
0
50
100
abs threshold(r= -0.57 N=30)
100
50
0
0
50
100
abs threshold(r= -0.78 N=28)
100
100
0
50
100
abs threshold(r= -0.68 N=16)
50
0
50
100
abs threshold(r= -0.29 N=30)
100
IFMC depth
100
0
IFMC depth
TMC slope
0
50
100
abs threshold(r= -0.59 N=24)
50
0
50
100
abs threshold(r= -0.80 N=9)
IFMC depth
IFMC depth
0
50
100
abs threshold(r= -0.32 N=23)
100
100
50
0
0
50
100
abs threshold(r= -0.51 N=19)
IFMC depth
0
100
50
0
0
100
0
0
50
100
TMC slope (r= 0.54 N=23)
100
50
0
0
50
100
TMC slope (r= 0.33 N=25)
100
50
0
0
50
100
TMC slope (r= 0.53 N=23)
100
0
50
100
abs threshold(r= -0.88 N=9)
50
TMC slope
50
100
0
50
100
abs threshold(r= -0.92 N=17)
50
0
IFMC depth
50
0
IFMC depth
0
50
100
abs threshold(r= -0.54 N=8)
50
IFMC depth
100
IFMC depth
IFMC depth
0
TMC slope
TMC slope
50
IFMC depth
TMC slope
TMC slope
100
TMC slope
100
50
0
0
50
100
TMC slope (r= 0.87 N=13)
50
0
0
50
TMC slope
100
Left/ right comparison
When a participant has data from both ears, they are included in a special scatter diagram plotting
left ear average statistics against the right ear. This example uses all impaired profiles but excludes
all participants with unilateral impairments.
TMC slope
IFMC depth
60
40
35
50
30
right ear
right ear
40
30
20
10
0
r= 0.94
N= 17
0
20
40
left ear
60
25
20
15
10
r= 0.94
5
N= 17
0
0
20
left ear
40
Paper copy (publish)
When this program is used in conjunction with MATLABs ‘publish’ facility, it can generate a .doc file
for further scrutiny. This will create a document containing each profile chart as well as the summary
statistics. The condition (e.g. ‘impaired’) needs to be set inside the plotAllFiles code. You will see that
a special area of code near the top of the program helps you to do that.
The name of the document file cannot be set in the publish command. By default it will be called
‘plotAllFiles.doc’. Rename the document file immediately after the run. This sequence pasted into
the command line will create the document file in a folder called ‘publishFiles’
91
options.outputDir='publishFiles';
options.format='doc'
options.showCode=false;
publish('plotAllFiles', options)
92
compareTwoProfileFolders
The following function uses plotAllFiles twice to compare the files in two different folders. In this
example all profiles from participants with normal hearing are grouped in a folder called
‘normalHearing’. The remainder are grouped in a folder called ‘impaired hearing.
compareTwoProfileFolders('normalHearing', 'impairedHearing')
This will have the same effect as scanAllFolders applied separately but also produces a summary
figure that compares the two groups on the basis of the three measures. The error bars are standard
deviations, showing the spread of the original scores.
From this figure we can see that
1. the absolute thresholds do not overlap between the groups
2. the slopes of the TMCs do overlap considerably
3. the TMC functions are shallower for the impaired group
4. the TMC slopes do not change much with frequency
5. the IFMC depth estimates are different between the two groups only at frequencies above
1000 Hz and, even then, there is considerable overlap even at these frequencies.
93
ParticipantCompendium
A .mat file called participantCompendium.mat contains a structure consisting of all the participants
along with other biographical information as well as the data for both the left and right ear. Navigate
to the profiles folder and type in the command window
load participantCompendium
This loads a single structure:
participant =
1x103 struct array with fields:
number
impaired
initials
matScript
(exists)
iffy
(problems with this listener’s data)
leftEar
(exists)
rightEar
(exists)
male
tinnitus
birthYear
startTest
age
code
leftEarData (data structure)
rightEarData (data structure)
This structure was compiled from data found in the participantDetails Excel files and all of the
individual data files (see above).
Profile format
The individual profiles begin life as .m files that pass a structure back to the calling program. A
profile represents a single profile for a single ear. Two profile files are supplied, one for each ear,
even if only one ear is measured.
The format is shown here only for the benefit of those who wish to create new profiles. The
information is derived from the output of the multiThreshold software. At present, there is no
automatic logging of the output in this format. The transcription needs to be performed manually. A
fixed format is used and missing data are presented as NaN (not a number).
function x = profile_BCR_R
x.BFs= [250 500 1000 2000 4000 6000 8000]; % abs threshold tone frequencies
x.LongTone= [10.5 6.9 21.3 30.1 56.4 55.7 77.0]; % thresholds (dB SPL)
x.ShortTone=[26.1 23.7 24.5 38.5 62.3 64.3
NaN];
94
x.IFMCFreq= [ ...
250 500
1000
2000
4000
6000 8000]; % IFMC probe frequency
x.IFMCs=[
NaN 41.38
40.50
44.14
66.92
NaN NaN
NaN 34.92
27.38
38.93
63.07
NaN NaN
NaN 31.17
32.05
43.87
66.58
NaN NaN
NaN 31.41
32.67
46.04
71.00
NaN NaN
NaN 39.61
18.04
49.21
71.86
NaN NaN
NaN 36.07
30.40
56.52
70.17
NaN NaN
NaN NaN 45.50
59.14
63.16
NaN NaN
];
% IFMC masker levels at masked threshold
x.MaskerRatio=[
0.5 0.7 0.9 1
1.1 1.3 1.6
frequencies (relative to probe frequency)
x.IFMCs= x.IFMCs';
% NB transpose
];
% masker
x.Gaps= [0.01 0.02 0.03
0.04
0.05
0.06 0.07
0.08
0.09]; % gaps
x.TMCFreq= [...
250 500
1000
2000
4000
6000 8000]; % TMC probe frequencies
x.TMC= [
NaN 44.09
34.07
49.06
73.70
NaN NaN
NaN 51.61
33.27
49.83
75.05
NaN NaN
NaN 57.10
35.62
50.78
79.35
NaN NaN
NaN 54.77
34.13
55.46
77.98
NaN NaN
NaN 58.47
40.27
54.48
79.94
NaN NaN
NaN 57.27
37.92
56.56
80.86
NaN NaN
NaN 60.05
42.35
57.36
80.32
NaN NaN
NaN 63.49
39.96
59.52
84.33
NaN NaN
NaN 61.22
42.01
60.33
86.97
NaN NaN
];
% TMC masker levels at masked threshold
x.TMC = x.TMC';
% NB transpose
95