Delays and Growth Rates of Multiple TEOAE Components
Shawn S. Goodman, Ian B. Mertes, and Rachel A. Scheperle
Citation: AIP Conference Proceedings 1403, 279 (2011); doi: 10.1063/1.3658098
View online: http://dx.doi.org/10.1063/1.3658098
View Table of Contents: http://aip.scitation.org/toc/apc/1403/1
Published by the American Institute of Physics
Delays and Growth Rates of Multiple TEOAE
Components
Shawn S. Goodman, Ian B. Mertes, and Rachel A. Scheperle
Department of Communication Sciences and Disorders, The University of Iowa
Abstract. Bandpass-filtered transient-evoked otoacoustic emissions (TEOAEs) show multiple energy peaks with time delays that are invariant with level and growth rates that vary with delay and
stimulus level, suggesting that multiple generation mechanisms may be involved at moderate and
high stimulus levels. We measured delays and magnitude growths of multiple TEOAE energy peaks
and compared the results obtained from linear and nonlinear extraction methods. To test the hypothesis that early components are generated at the basal portion of the cochlea, delays and growth
rates were also measured in the presence of highpass masking noise for a subset of subjects. No
effect of the highpass masking was seen. The results are discussed in terms of potential generation
mechanisms of the multiple energy peaks.
Keywords: transient-evoked otoacoustic emissions, growth, delay, multiple components, generation mechanisms
PACS: 43.64.Jb, 43.64.Kc, 43.64.Ri
INTRODUCTION
Otoacoustic emissions (OAEs) may be composed of multiple components generated by
one or more mechanisms. Distortion-product (DP) OAEs have wave-fixed and placefixed components generated by distortion and reflection sources, respectively [9, 15].
Stimulus-frequency (SF) OAEs may be generated by an early component having a
slowly-rotating phase and a late component with a rapidly-rotating phase [11]. Transientevoked (TE) OAEs have been shown to have multiple energy peaks with differing time
delays and growth rates [1, 4]. Similar findings have been reported for DPOAEs [12]
and SFOAEs [2].
In humans, SFOAEs and TEOAEs are believed to be generated by place-fixed coherent reflection, at least at low stimulus levels [6, 8]. At higher stimulus levels, it is
possible that TEOAEs may have a distortion component [4, 13, 14].
The purpose of this study was to investigate multiple TEOAE components with regard
to time distribution, magnitude growth, and delay as a function of stimulus level. The
magnitude growth functions obtained using linear and nonlinear extraction methods
were compared. Potential generation mechanisms are discussed in relation to these
characterizations.
METHODS
Transient stimuli (1–4 kHz bandwidth, 2 ms duration) were presented to 18 normalhearing ears at 7 levels (45–75 dB peSPL, in 5 dB steps). Five ears had synchronous
spontaneous emissions, and were excluded from analysis. Presentation and recording
What Fire is in Mine Ears: Progress in Auditory Biomechanics
AIP Conf. Proc. 1403, 279-285 (2011); doi: 10.1063/1.3658098
© 2011 American Institute of Physics 978-0-7354-0975-0/$30.00
279
were made using an ER-10C (Etymōtic Research) probe assembly and custom MATLAB software on a PC. Stimuli were presented at a rate of 33.3 clicks per second, and
1000 averages were obtained for each stimulus level. Emissions were extracted using a
linear paradigm and a double-evoked nonlinear residual paradigm [7]. For the nonlinear
paradigm, suppressor clicks were presented at 12 dB above the probe clicks.
Averaged TEOAE waveforms were filtered with a 1/3-octave wide bandpass finite
impulse response filter centered at 2.52 kHz and corrected for group delay. Waveform
envelopes were calculated as the magnitude of the analytic signal
(1)
env = x2 + x̂2 ,
where x̂ is the Hilbert transform of the waveform, x. Envelopes were examined to locate
energy peaks. Time zero was defined as the peak of the stimulus. The analysis window
began at 2 ms in order to avoid stimulus artifact. The absence of stimulus artifact after 2
ms was verified in an artificial ear and a subject with severe sensorineural hearing loss.
Three ears were also tested with the addition of ipsilateral highpass masking noise.
For this condition, the bandwidth of the 2 ms transient stimulus was reduced to 1–3
kHz and was presented at 75 dB peSPL. Masking noise (3.2–20 kHz bandwidth) was
mixed acoustically with the transient stimulus. The level of masking noise was adjusted
until a 2 f 1 − f2 DPOAE at 2.5 kHz ( f2 = 4 kHz, f2 / f1 = 1.22) was reduced by at least
6 dB while maintaining a 12 dB or greater signal-to-noise ratio. This noise level was
used during presentation of transient stimuli. Analysis of these TEOAE waveforms was
performed using the same analysis described above.
RESULTS
Figure 1 shows the envelopes obtained for three subjects. Multiple energy peaks
are present in all three subjects. Within subjects, peak delays were nearly constant across stimulus level, allowing construction of a level series for each set
of peaks having similar delays. In contrast to the stability of individual peaks
across level, the delay location of the peaks was variable across subjects.
Magnitude (µPa)
400
NH000R
NH010L
NH003L
300
200
100
0
2
3
4
5
6
7
8 2
3
4
5
6
7
8 2
3
4
5
6
7
8
Delay (ms)
FIGURE 1. TEOAE envelopes (linear extraction mode). Each panel shows data from a different
subject. Each envelope corresponds to one of seven stimulus levels (45–75 dB peSPL). TEOAE
peaks, identified using a peak-picking algorithm, are shown by small dots. Level series were constructed for each energy peak for all ears and are plotted in Fig. 2.
280
Normalized Magnitude (dB)
The magnitudes of early peaks grew
35
faster with stimulus level than later
30
peaks.
Figure 2 shows the level series of all
25
ears tested. Each level series is repre20
sented by a single line. Peak magnitudes were expressed in dB SPL and
15
normalized so that the lowest magni10
tude was zero. Most of the lines are
5
nearly vertical, indicating constant delay across stimulus level. Although a
0
4
5
6
7
8
2
3
few of the lines have a slope, indicatDelay (ms)
ing increasing or decreasing delay with
stimulus level, no systematic trend
FIGURE 2. Normalized peak magnitudes and
delays (25 series, 13 ears) obtained by the linwas seen. Given that each line shows
ear extraction paradigm. Each level series is repthe peak magnitudes across all seven
resented by a single line. Small filled circles (obstimulus levels, shorter lines represent
tained from the peak picking operation shown in
more compressive growth than longer
Fig. 1) on the lines show the magnitude at each of
lines. The orderly arrangement of line
the seven stimulus levels. The gray shaded region
length demonstrates that growth beshows the 95% confidence interval of expected
SFOAE latencies [10].
comes increasingly compressive with
longer delays. Note that peak delays
are spread evenly across the range of 2–8 ms, with the exception of a gap at 7 ms and
a smaller gap at 3.5 ms. 7 ms is the expected upper limit of SFOAE delays for 2.5 kHz
tones [10]. Peaks with delays longer than 7 ms likely represent multiple internal reflections. It is uncertain whether a true gap exists around 3.5 ms. If so, it might separate
peaks into “early” and “late” components; however, more data might show the distribution to be essentially continuous.
A quantitative description of growth rates is shown in Fig. 3. Growth rates were
determined as follows: Each level series was fit with an exponential function of the
form
(2)
y = aebx + cedx ,
where x is stimulus level in dB peSPL. The resulting fits were differentiated to obtain
the growth rates (dB per dB). The growth rates of the individual level series were then
used to obtain the fits shown in Fig. 3. For each stimulus level, growth rates as a function
of peak delay (top right panel) were determined using an exponential fit of the form
y = aebx ,
(3)
where x is delay in ms. For integer delays between 2 and 8 ms, growth rates as a function
of stimulus level were determined using an exponential fit [Eq. (2)]. Combining these
two functions (bottom panel) shows that growth is most compressive for peaks with long
delays elicited by high stimulus levels. Growth is expansive for peaks with short delays
elicited by low stimulus levels.
Previous work has described growth rates for TEOAEs extracted using a nonlinear
paradigm [4]. Because nonlinear paradigms cancel any linear growth of the TEOAEs,
281
Growth Rate (dB/dB)
1.4
1.2
2 ms
45 dB
1.0
0.8
0.6
75 dB
0.4
8 ms
0.2
0
45
50
55
60
65
70
75
2
3
4
5
6
7
8
Delay (ms)
Stimulus Level (dB peSPL)
1.4
Growth Rate (dB/dB)
1.2
1.0
0.8
0.6
0.4
0.2
0
2
4
6
Delay (ms)
8
75
70
65
60
55
50
45
Stimulus Level (dB peSPL)
FIGURE 3. Growth rates as a function of peak delay and stimulus level (linear extraction mode).
Linear growth is indicated by a value of 1. The top panels display growth rates as a function of
stimulus level (left) and peak delay (right). Each line in the top left panel represents a peak delay (2,
3, 4, 5, 6, 7, or 8 ms, from top to bottom). Each line in the top right panel represents a stimulus level
(45, 50, 55, 60, 65, 70, and 75, from top to bottom). The bottom panel shows the data combined in
a three-dimensional surface plot.
it was of interest to compare growth rates obtained using both nonlinear and linear paradigms. Figure 4 shows a comparison of growth rates obtained using the two
paradigms. Overall, growth rates were higher when using nonlinear extraction. At high
stimulus levels, nonlinear extraction showed more expansive growth of early peaks,
while both extraction methods showed similar compressive growth for later peaks. At
lower stimulus levels, the growth rates of early peaks were similar between the two
methods. The constant expansive growth across time demonstrated by the nonlinear extraction at the lowest stimulus level may be an artifact arising from poor signal-to-noise
ratio. Taken together, these data show that extraction method has the largest effect on
growth rates for peaks early in time elicited by higher stimulus levels.
282
DISCUSSION
Our data show the following:
75 dB
Growth Rate (dB/dB)
1. TEOAEs at a given frequency region are composed of multiple energy peaks.
2. Across ears, these peaks appear to
be distributed nearly continuously
in time.
3. Within ears, peak delays are generally constant across stimulus
level.
4. Growth rates of peaks decrease
(growth is more compressive)
with increasing delay and increasing stimulus level.
70 dB
65 dB
60 dB
55 dB
TEOAEs, in a manner similar to
SFOAEs, are thought to be generated
50 dB
by a coherent reflection mechanism
45 dB
located near the peak of the traveling wave [6]. TEOAE peaks occurring
early in time, however, must be gener2
3
4
5
6
7
8
Time
(ms)
ated from a different place and/or by
a different mechanism. As a prelimiFIGURE 4. Growth rates obtained using linear
nary investigation into the generation
(thin lines) and nonlinear extraction (thick lines).
of early peaks, we repeated our meaThe stimulus level for each set of responses is
surements in three subjects with the
shown. Dotted lines represent linear growth.
addition of highpass masking noise.
The presence of masking did not affect
the magnitudes or delays of the peaks compared to responses obtained without noise.
This suggests that the early components were not influenced by 2 f 1 − f2 distortion products (see Methods section), though it does not rule out the presence of other distortion
sources. We also calculated group delay for early and late peaks (not shown). Although
the phases of early peaks rotated less rapidly than late components, rotations were consistent with the peak delays generated by a reflection source, which is not consistent
with a wave-fixed distortion generation mechanism.
It has also been suggested that early components arise from basal sites in the tail
region of the traveling wave. Our results do not support that early peaks arise from the
most basal regions of the cochlea. A schematic of the masking paradigm is shown in
Fig. 5. A possible explanation for the generation of early TEOAE energy peaks is placefixed reflections from locations somewhat basal to the peak. The expected lower limit
of round-trip delay for a 2.5 kHz SFOAE was 5.13 ms [10]. Using a cochlear model
from Geisler [3], the number of cycles to the peak of the traveling wave was 5, and the
number of cycles to the location corresponding to the masking cutoff frequency was 2.5.
Travel times were approximated by dividing the total number of cycles to the peak by
283
the forward SFOAE delay, yielding
Characteristic Frequency (kHz) 3.2 2.5
13 14.7
Distance from footplate (mm)
the delay per cycle. Because components generated basal to the peak
of the traveling wave will have gone
Masking Noise
through fewer phase rotations, their
travel times can be much faster than
components generated very near the 2.5
peak. In this scenario, it is possible for 5.1
a 2.5 kHz component being reflected Delay (ms)
from the 3.2 kHz place to have a
FIGURE 5. Schematic of masking condition.
round-trip travel time of approximately
Traveling wave shown for 2.5 kHz stimulus. Mask2.5 ms, which is consistent with our
ing noise is shown by the gray box. Distance from
data. The increasingly linear growth of
stapes calculated using Greenwood map [5].
earlier components is also consistent
with generation sites basal to the peak of the traveling wave.
ACKNOWLEDGMENTS
Supported by a grant from the American Speech-Language-Hearing Foundation.
REFERENCES
[1]
Carvalho S, Buki B, Bonfils P, Avan P (2003) Effect of click intensity on click-evoked otoacoustic
emission waveforms: Implications for the origin of emissions. Hear Res 175: 215–225
[2] Choi Y, Lee S, Parham K, Neely ST, Kim DO (2008) Stimulus-frequency otoacoustic emission:
Measurements in humans and simulations with an active cochlear model. J Acoust Soc Am
123:2651–2669
[3] Geisler CD (1998) From Sound to Synapse: Physiology of the Mammalian Ear. New York: Oxford
University Press
[4] Goodman SS, Fitzpatrick DF, Ellison JC, Jesteadt W, Keefe DH (2009) High-frequency clickevoked otoacoustic emissions and behavioral thresholds in humans. J Acoust Soc Am 125:1014–
1032
[5] Greenwood DD (1990) A cochlear frequency-position function for several species–29 years later. J
Acoust Soc Am 87:2592–2605
[6] Kalluri R, Shera CA (2007) Near equivalence of human click-evoked and stimulus frequency
otoacoustic emissions. J Acoust Soc Am 121:2097–2110
[7] Keefe DH (1998) Double-evoked otoacoustic emissions: I. Measurement theory and nonlinear
coherence. J Acoust Soc Am 103:3489–3498
[8] Knight RD, Kemp DT (1999) Relationships between DPOAE and TEOAE amplitude and phase
characteristics. J Acoust Soc Am 106:1420–1435
[9] Shera CA, Guinan JJ (1999) Evoked otoacoustic emissions arise by two fundamentally different
mechanisms: A taxonomy for mammalian OAEs. J Acoust Soc Am 105:782–798
[10] Shera CA, Guinan JJ, Oxenham AJ (2002) Revised estimates of human cochlear tuning from
otoacoustic and behavioral measurements. Proc Natl Acad Sci USA 99:3318–3323
[11] Shera CA, Tubis A, Talmadge CL (2008) Testing coherent reflection in chinchilla: Auditory-nerve
responses predict stimulus-frequency emissions. J Acoust Soc Am 124:381–395
[12] Stover LJ, Neely ST, Gorga MP (1996) Latency and multiple sources of distortion product otoacoustic emissions. J Acoust Soc Am 99:1016–1024
284
[13] Withnell RH, Hazlewood C, Knowlton A (2008) Reconciling the origin of the transient evoked
otoacoustic emission in humans. J Acoust Soc Am 123:212–221
[14] Yates GK, Withnell RH (1999) The role of intermodulation distortion in transient-evoked otoacoustic emissions. Hear Res 136:49–64
[15] Zweig G, Shera CA (1995) The origin of periodicity in the spectrum of evoked otoacoustic emissions. J Acoust Soc Am 98:2018–2047
COMMENTS AND DISCUSSION
Christopher Shera: A very nice analysis. Any thoughts on how to understand the gap at 7 ms? That
is, thoughts on why there should never be a maximum in the time-domain waveform near the expected
SFOAE delay time?
Sarah Verhulst: Have you investigated the spectra of the waveforms after you band-pass filter the
recorded TEOAE with the 1/3rd octave wide filter? I am wondering whether there was only one frequency
component in those waveforms or whether there were multiple evoked components. In case of multiple
evoked components, have you considered a possible relation between the beating frequency of the evoked
components and the delay of the envelope peaks?
285