RESUMEN DE LOS FACTORES POTENCIALES Y LIMITANTES DE

The effect of a third tone on DPOAE at intermediate ratios of ^/fi is very unpredictable and different patterns can be achieved with only slightly different stimulus parameters. It is not possible to determine the DP sources from the detailed patterns seen with such stimulus configurations.

Although it is difficult to make detailed conclusions, the effects on DPOAE phase are

consistent with the existence o f two sources of 2î\-ïi DPOAE, which may be in or out

Mechanisms o f otoacoustic emissions F : Wave and place fixed DPOAE maps 130

F. Separate wave and place fixed DPOAE maps of human ears

I Introduction

In chapter D, detailed DPOAE measurements were presented and discussed. It was observed that measured DPOAEs took one of two forms which, it is proposed, are a result o f either a wave fixed or a place fixed emission mechanism. Emissions from both mechanisms may occur simultaneously, but whichever DP emission path results in the higher DP amplitude in the ear canal will dominate the phase behaviour o f the measured DPOAE. The dominant emission mechanism for the Zfr-f] DP is primarily determined by the stimulating frequency ratio fiz/fi. This dependence on the frequency ratio was first reported by Kemp (1986).

The results in section D also showed that the pattern of place fixed behaviour o f the Zfi-f] DP when f^/fi is small is contiguous with the pattern that is seen in the 2f2-fi DP

for all fi/fr. However the ear canal DPOAE was dominated by a wave fixed

emission mode when fr/fi exceeded approximately 1.1. A similar relationship was also

seen between the 3fi-2fr and DP’s, but with the transition between the wave

fixed and place fixed emissions occurring at a smaller frequency ratio.

a) Reasons to separate the wave and place fixed emissions

Whilst the data in section D are interesting and informative, the full picture o f the wave and place fixed emissions cannot not seen because the DPOAE behaviour that is observed is dominated by whichever mode is the stronger. Therefore it is not known whether the two DP paths both continue to be significant for all fg/fi ratios or whether DP production tends to be bi-modal with only one mode usually present and just a small transition region at which both emissions can occur. If both modes were always present and could be separated then the phase behaviour o f the ‘residual’ or ‘minority’ emissions could be studied in the context of the travelling wave excitation patterns. The systematic trends that have previously only been seen in restricted stimulus parameter conditions could be pursued over a wider range of frequency ratios.

Observation of the place fixed phase behaviour will be of particular interest, especially through the crossover between the lower and upper sidebands. The lower sideband f^/fi ratio at which the phase pattern which continues from the upper sideband changes, and the phase behaviour in the lower sideband at moderately wide ratios would also be

Mechanisms o f otoacoustic emissions F : Wave and place fixed DPOAE maps 131 interesting. These features will provide further clues regarding the locations o f the initial DP generating region relative to the 1% frequency place and the reflecting region relative to the fbp place.

b) Methods to separate the wave and place fixed DPOAE components

Separating the wave and place fixed emissions is not a straightforward task. One approach is to add a third stimulus tone to suppress activity at the DP place, leaving only the DP emitted directly from the fi region. Dreisbach and Siegel (1999) have proposed a method in which vector subtraction of DPOAE measurements obtained with and without the third tone can yield the DP place contribution. However Siegel et al (2000) have shown that it can be difficult to ensure that the DP place is fully suppressed without the fz region, and hence DP generation, also being affected. This risk is greatest with small stimulus frequency ratios.

An alternative method is pursued here, in which the wave and place fixed DPOAE are separated by a phase gradient dependent post processing method. Shera and Zweig (1993) and Zweig and Shera (1995) have used Fourier transforms to investigate the frequency domain of SFOAE. Stover et al (1996a) and Fahey and Allen (1997) used Fourier transforms on DPOAE frequency sweeps to observe the phase gradient-derived latency.

This method has the advantage of avoiding the unpredictable nonlinear interaction effects of a third tone. For definite separation of emissions with different phase gradient behaviours to be achieved, a long frequency sweep is required in order for the Fourier transformations to be of sufficient resolution. This technique has the potential to allow the ‘wave’ and ‘place’ fixed DPOAEs to be observed separately, allowing behaviour which is usually masked to be uncovered.

There is a fundamental difference between the third tone (suppression) method and the Fourier transform method. The third tone selectively affects the emission modes on the basis o f their place on the basilar membrane and it turns out that these emissions are different modes. The Fourier transform method separates the emission modes on the basis o f their different phase gradients and their probable origin from different places in the cochlea is not relevant to the technique. The suppression and phase gradient methods are comparable if the ‘place fixed’ emission happens to be from the region of the DP frequency place and the ‘wave fixed’ emission is from elsewhere.

In the present study, inverse Fourier transforms are applied to constant f^/fi DPOAE frequency sweeps. Wave and place fixed DPOAE have different phase gradients with

Mechanisms o f otoacoustic emissions F: Wave and place fixed DPOAE maps 132 these frequency sweeps, so in the ‘time’ domain peaks are produced at different time delays. Low and high latency DPOAE components are then windowed in the quasi-time domain and then transformed back to the frequency domain separately so that new separate maps can be drawn.

n M easurements and Data Analysis

Data presented here have been obtained from an additional analysis o f data from subjects RK and RN in figures 27 and 28. The DPOAE measurements which have been used for this analysis were obtained with Li=L2=70 dB SPL with DP frequencies ranging from 1 kHz to 4.1 kHz in 12 Hz steps. Stimulus frequency ratios from 1.01 to 1.5 were employed. From these data the two components o f interest, which have differing latencies, were separated by the following post-processing method.

A 512 point frequency array was set up, ranging from 0-3100 Hz, thereby preserving the original 3100 Hz range. An exponential frequency transformation was adopted in order to linearise the underlying curve in the phase versus frequency relationship o f the place fixed DPOAE and therefore produce clearer peaks in the time domain. The exponential frequency points were calculated as follows:

fi=1000x4.1^^'''^^^” ^-1000

.. where fi=the frequency of the i th array point.

For convenience only, the measured data points, which span 1000 Hz -4100 Hz, were frequency shifted down 1000 Hz and inserted into this array with linear interpolation between data points where required. The level and phase data were converted to complex number format for the inverse Fourier transformation.

No frequency domain windowing was applied before the inverse Fourier transformation. Although there is a risk of artefacts, the artefacts are easily distinguishable from the signal in this case as the artefacts would be spread throughout the time window whereas the true signal would form peaks early in the time window. The time domain windowing which was subsequently employed would therefore largely remove any such artefacts. The positive advantage of this approach was that the data were not weighted in favour of data from the middle o f the frequency range and the final separated data were more easily compared with the original unprocessed data.

Mechanisms o f otoacoustic emissions F: Wave and place fixed DPOAE maps 133 As the frequency array had an exponential characteristic, the meaning o f the time scale o f the inverse Fourier transform was not straightforward. The timescale in the time domain which is indicated in figure 41 was calculated as follows:

Time increment per point (ms)= 1000

frequency range(Hz)

This gives a time increment of approximately 0.32 ms per point.

This calculated time interval most accurately reflects true time for data from the middle of the frequency range (2-3 kHz). It is an underestimate for the lower frequencies and an overestimate for the higher frequencies because o f the exponential frequency transformation used. The definition of this timescale is not relevant for the subsequent windowing and processing of the data. The transformation into the ‘time’ domain produced separate amplitude peaks representing the low and high latency components (figure 41).

The peaks were separated by time-windowing. The windowing was designed to separate the wave and place fixed DPOAE and the crossover point was set at the ‘time’ defined midway between the wave and place fixed peaks. The chosen crossover point for the ‘early’ and ‘late’ time windows was the 5^ data point, which is about 1.6 ms. A flat topped window with a raised cosine edge was adopted. The cosine curve ‘edge’ extended from the 3"^^ to the 7^ data points (approx. 1.00-2.25 ms). The ‘late’ window was rolled off more gradually between the 40^ and 50^ data points (12.9-16.1 ms), comfortably after the DPOAE peaks had died down to the noise floor.

Each windowed part of the time domain response was returned to the frequency domain by a forward Fourier transform. The data were returned to level and phase format and converted back to the original 256 point linear frequency array.

Mechanisms o f otoacoustic emissions F: Wave and place fixed DPOAE maps 134