Download Models of signal processing in human hearing

Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
www.elsevier.de/aeue
Models of signal processing in human hearing
Roland Sottek∗ , Klaus Genuit
HEAD acoustics GmbH, Research NVH, Ebertstr. 30a, Germany
Received 16 March 2005
Dedicated to Prof. em. Dr.-Ing. E. h. Hans Dieter Lueke on the occasion of his 70th birthday
Abstract
The paper presents a theoretical system model of the monaural signal processing in the human auditory system. Moreover,
it deals with the application of this hearing model to the calculation of just-noticeable amplitude differences and variations,
as well as elementary psychoacoustic parameters and a deconvolution algorithm for the detection of dominant spectral components based on the excitation pattern of the basilar membrane. The use of the hearing model for sound quality tasks allows
evaluating the temporal and spectral patterns of a sound (“Relative Approach” analysis) where absolute level or loudness is
often without significance.
䉷 2005 Elsevier GmbH. All rights reserved.
Keywords: Hearing model; Auditory discrimination capabilities; Psychoacoustic parameters; Deconvolution algorithm; “Relative
Approach” analysis
1. Introduction
The performance and versatility of human hearing as a
sound analyzer remain unmatched by technical systems. In
general, physically oriented analytical methods do not compare with humans in determining sound quality. These analytical methods must be complemented by psychoacoustic
models to consider the sensation of humans.
The normal way to develop a mathematical model of
sound perception is to look at the relations between acoustic stimuli and sensations. Often the result is an empirical
formula which can describe only a few aspects of sound perception. Because there are so many models it is very difficult
to get an idea of the complete signal processing in human
hearing. Thus it would be desirable to have one principle
of overriding importance, which covers many psychoacoustic phenomena. Based on the physiology of the ear, a
∗ Corresponding author. Tel.: +49 2407 577 61; fax: +49 2407 577 99.
E-mail address: [email protected] (R. Sottek).
1434-8411/$ - see front matter 䉷 2005 Elsevier GmbH. All rights reserved.
doi:10.1016/j.aeue.2005.03.016
theoretical system model has been developed with the intention of explaining and describing psychoacoustic effects
and basic auditory sensations [1].
The main work has been to create an ear-related
time–frequency representation. This time–frequency analysis is based on the separation of acoustic signals into
bandpass signals in connection with a new deconvolution
algorithm.
It can be shown that the vibration pattern of the basilar membrane may be represented as the convolution of
the physical spectrum with an ear-related function, with the
proviso that an adequate frequency transform is used. The
neural processing in the brain is considered to be a highresolution spectral analysis that looks for dominant spectral
components and supplies them to a recognition process.
One of the remarkable features of this hearing model is
its easy mathematical handling and good correlation with
the results of many psychoacoustic experiments. For example, just-perceptible amplitude and frequency variations
have been predicted. In this case the nonlinear processing of
158
R. Sottek, K. Genuit / Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
human hearing has a fundamental meaning. Another main
emphasis lies in the calculation of basic auditory sensations
such as the psychoacoustic parameters loudness, roughness
and fluctuation strength.
2. Functional block diagram of the hearing
model
The basic processing steps of the hearing model are shown
in Fig. 1. The pre-processing consists of filtering with the
outer ear–inner ear transfer function. A large number of
asymmetric filters (with a high degree of overlapping) model
the frequency-dependent critical bandwidths, and the tuning curves, of the frequency-to-place transform of the inner ear which mediates the firing of the auditory hair cells
as the traveling wave from an incoming sound event progresses along the basilar membrane. The inconstant percentage bandwidth versus frequency of the auditory filter bank
conveys a high-frequency resolution at low frequencies and
a high time resolution at high frequencies with a very small
product of time and frequency resolution at all frequencies,
which empowers, e.g. human hearing’s recognition of shortduration low-frequency events. Subsequent rectification accounts for the fact that the nerves fire only when the basilar
membrane vibrates in a specific direction. The firing rates
of the nerve cells are limited to a maximum frequency. This
Human Physiology
is modeled using low-pass filters. A feature of the hearing
model in its usual psychoacoustic application to human perception is a compressive nonlinearity relating sound pressure
to perceived loudness. The hearing model spectrum versus
time (Fig. 2), however, is accessible as an intermediate result before implementation of the compressive nonlinearity,
making the algorithm’s time/frequency capabilities available
as a linear tool comparable to conventional techniques for
events at sound pressures beyond normal human hearing
perception limits.
Based on these fundamental processing steps there are a
few post-processing mechanisms for the calculation of the
basic auditory sensations, or the already-mentioned deconvolution algorithm (not shown in the functional diagram,
thought to work at a higher neural level). As one example
the loudness results are obtained by summing all specific
loudness signals.
3. Design of the ear-related filter bank
3.1. Frequency scales
The nonlinear frequency-to-place transform f = y(z) can
be deduced directly from the frequency-dependent bandwidth of the auditory filters, assuming that the bandwidth
f (f ), independent of the actual frequency, always corresponds to a section of constant length on the linearly divided
Hearing Model
Outer and middle ear filtering (preprocess)
Outer and middle ear
Basilar membrane, traveling wave
Hair cells, firing rates
Brain, neural processing
Ear-related filter bank using band-pass filters which
are implemented using equivalent low-pass filters;
formation of envelope, consideration of the
frequency-dependent threshold in quiet by a
reduction of excitation level
The nerve cell firing rates follow the signal magnitude
curve only up to a maximum frequency, modeled in a
simplified manner by a 3rd-order-low-pass filter. In an
additional step, the individual bandpass signals are
processed using a nonlinearity as a function of the
level of the exciting signal.
Calculation of various sensations, such as specific
loudness, or effects like tonal deconvolution:
Other neural processing of the brain follows in the
human hearing system, resulting in the different
sensations such as loudness, roughness,
impulsiveness, and pitch detection. Specific
mathematical algorithms, such as spectral
deconvolution in the case of pitch detection, are
employed to describe these effects.
Fig. 1. Steps of the human hearing process (top to bottom).
R. Sottek, K. Genuit / Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
159
Time Signal
Channel 1
a1
Auditory
Sensation
Channel i
ai
Channel n
Specific Roughness
Specific Loudness
an
Specific Fluctuation
Outer and
Middle Ear
Filtering
Ear-related Filterbank
Formation of Envelope
Lowpass
Nonlinearity
20 Bark
15
10
5
0
0
0.1
0.2
0.3
0.4 s
Aurally-adequate Spectral Analysis
of a Door Slam Noise
Fig. 2. Functional block diagram of the hearing model, showing the formation of the hearing model spectrum versus time as an intermediate
result. Based on this result a compressive nonlinearity models the relationship of sound pressure to perceived magnitude in psychoacoustic
measurements. Frequency scaling is shown in Bark (critical band number).
basilar membrane (z-scale). The frequency f results from
integration over the bandwidth function f (f ), which must
be expressed as a function of the basilar membrane coordinates z
z
f [y(z̃)] dz̃.
(1)
y(z) =
0
We are looking for f = y(z) or the inverse function
z = y −1 (f ). The solution of Eq. (1) is given (with an arbitrary integration constant C) by
1
z=
df + C.
(2)
f (f )
As an example,
f (f ) = f (f = 0) + c · f
(3)
can be used as an approximation to the well-known Bark
scale [2] or the ERB (equivalent rectangular bandwidth)
scale [3] (with different constants f (f = 0) and c). Using
this bandwidth function, Eq. (2) can be rewritten as
z=
1
c
ln
f +1 .
c
f (f = 0)
(4)
Hereby, the constant C has been calculated to achieve
y(0) = 0. The frequency is given by
f=
f (f = 0)
[exp(c · z) − 1].
c
(5)
It should be noted that Greenwood describes the frequencyto-place transform of the basilar membrane based on physiological measurements of v. Békésy using the following
160
R. Sottek, K. Genuit / Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
Fig. 3. Excitation on the basilar membrane due to individual tones; left: linear-frequency scale, right: hearing-related ERB scale.
formula (details in [4]):
f
z = 18.31 · lg 6.046
+1 .
kHz
Substituting f (f ) and y(z) with the Eqs. (3) and (5) yields
(6)
Comparing Eq. (6) with Eq. (4) yields for the Greenwood
approximation: c ≈ 0.126 and f (f = 0) ≈ 20.8 Hz.
3.2. Shape of the auditory filters
The impulse responses of the auditory band-pass filters
have been technically implemented as modulated low-pass
filters1
hF (t) = 2hT ,F (t) exp(j2F t)
using the low-pass function2
1
t
1 1 n−1
hT ,F (t) = (t)
exp −
(n − 1)! with the corresponding Fourier transform
f
1
HT ,F (f ) =
=
H
,
T
[1 + j2f (F )]n
f (F )
(7)
(8)
(9)
where n is the filter order3 and (F ) is a time constant,
related to f (F ) by
1
1
2n − 2
f (F ) = 2n−1
.
(10)
n − 1 (F )
2
The spectrum of one specific output signal of the filter bank
is then given by
GF (f ) = S(f ) · HT ,F (f − F ).
(11)
The output signals of the whole filter bank for a sinusoidal
input with frequency fi and amplitude s are
fi − F
y(zi ) − y(z)
Efi (F ) = s ·HT
= s ·HT
.
f (F )
f [y(z)]
(12)
1 The magnitude of the corresponding transfer functions agrees well
with the ROEX-filters defined by Patterson [5]. Whereas Patterson defined
the magnitude only (which is sufficient to determine stationary excitation
patterns in the spectral domain), the implemented minimum-phase filters
allow processing in the time domain [1].
2 (t) is the unit step function.
3 Filter order n = 5 is recommended.
Efi (F ) = s ·HT
exp[c · (zi − z)] − 1
c
= s ·HT (zi − z).
(13)
We have seen that the shape of the filter bank output for
a sinusoidal input does not depend on the frequency if a
hearing-related frequency transform according to Eq. (5) is
used (Fig. 3). Only a shift in z occurs. The output can be
considered in this case as a convolution of the physical spectrum with a “characteristic function” E0 (z) – the typical
excitation pattern of a sinusoid (shifted to z = 0). Even in
the case of more complex input signals like a sum of sinusoidal components with time-varying amplitudes, frequency
and phase relationship, this statement remains valid.
The following example (Fig. 4, upper figure) shows first
the use of the hearing model spectrogram versus time for a
speech signal (taken from [1]). The output of the filter bank
is interpreted as a spectral representation of the given signal
blurred by the spectrum of the “window” E0 (z). A deconvolution algorithm has been applied to the excitation pattern
for the detection of dominant spectral components (lower
figure). This algorithm works as an iterative procedure, detecting dominant spectral components (as described in [6])
and estimating magnitude and phase of these spectral lines
while considering the influence of the blurring window.
The basilar membrane coordinate is shown in Fig. 4 on the
left-hand side. Zero millimeter corresponds to the starting
point at the oval window (location of highest frequencies).
Thirty-two millimeter (length of the human basilar membrane) corresponds to the apical end of the membrane (lowest frequencies). The corresponding frequencies are shown
on the right-hand side. The highest level is indicated by black
and the lowest level by white. The time delays between different channels have been compensated, e.g. the maxima of
the impulse responses of the filters occur at the same time.
Fourteen channels per mm basilar membrane length have
been used for the calculation. The fundamental frequency of
the speech signal (approximately 100 Hz) and the harmonics can be seen in both diagrams, although the lower figure
shows a much higher resolution and a very clear representation of the signal.
R. Sottek, K. Genuit / Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
161
Fig. 5. Level increment (solid line) and decrement (dashed line)
L of a 1 kHz tone, required for double and, respectively,
half-loudness perception. The diagram shows the mean values of
12 test subjects and the range containing 50% of the measurements
(according to Zwicker [7]).
Fig. 4. Excitation pattern of the basilar membrane (hearing model
spectrum versus time) due to a speech signal: the German word
“allein” (upper figure) and the result of the search for tonal components via deconvolution algorithm (lower figure).
3.3. Nonlinear signal processing
The nonlinearity of the auditory system is especially significant for the loudness perception. The specific loudness
distribution, resulting from the application of this nonlinearity to the excitation pattern, also forms the basis for calculating other psychoacoustic parameters such as roughness
or fluctuation4 (Fig. 2). Moreover, the nonlinearity mainly
determines the just-perceptible amplitude and frequency
modulations.
Psychoacoustic measurements were performed by
Zwicker to determine the level dependence of loudness for
tonal signals, in which test persons had to set the level in
such a way that double, or half the initial loudness was
perceived (Fig. 5). The mean value of the set-level lies between 10 and 18 dB for a reduction by half, and between
7 and 14 dB for a doubling. At a certain initial level, the
scatter range for level reduction is up to 12 dB and at least
6 dB. The higher scatter for level reduction is explained by
4 Fluctuation and roughness calculations only differ in frequency
weighting: a band-pass filter instead of a high-pass filter is applied to the
envelope variations.
the influence of a previous tone on the capacity to evaluate
subsequent tones. For simplification, a 10 dB level change
for both doubling and halving of loudness has been chosen, which corresponds to a power law with an exponent
of ≈ 0.6 regarding total loudness. The exponent of the
power law has been set to = 0.5 for the calculation of
specific loudness values in ISO 532B. Data published by
Stevens also show enormous scatter [8].
Due to the considerable scatter it must be doubted whether
the test subjects are at all in the position to set the loudness
(or other perception related parameters) at certain ratios (for
discussion of problems related to loudness scaling, refer to
[9,10]).
Garner [11] developed a procedure which constructs the
“true” chosen ratio of loudness values (instead of the desired
ratio of two) finally showing a much lower exponent ≈
0.35. It is also noteworthy that the exponent found by Garner
agrees very well with data obtained by measurements of the
transverse velocity of the basilar membrane (0.2 < < 0.3)
[12].
It is much easier for test subjects to decide whether two
signals generate different or equal sensations, than for them
to adjust levels to double the loudness of a reference signal. Following this approach various detection experiments
have been studied in [1] assuming that our hearing system
analyzes specific loudness in equidistant sensation steps.
For e.g. presenting two tones with sound pressures p1 and
p2 , a level change is detected, if the maximum specific
loudness produced by these tones differ at least by N ,
i.e.
p2 = y −1 [y(p1 ) + N ],
(14)
y is a nonlinearity function, y −1 is the inverse function and
N is the equidistant quantization step of the sensation
(specific loudness) derived from several data published in
162
R. Sottek, K. Genuit / Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
Fig. 6. Just-noticeable amplitude differences L for
quasi-stationary tone impulses (frequency: 1 kHz). The data taken
from Houtsma et al. (open circles, from [13]) and Schorer (filled
circles) [13] can be best approximated for levels between 15 and
80 dB using a power law with the exponent = 0.2.
the literature. If we plot the level differences
Fig. 7. Just-noticeable amplitude modulation at f = 1 kHz and
sinusoidal modulation with fmod = 4 Hz, data from measurements
published by Riesz (open circles, from [13]) and Schorer (filled
circles) [13] as well as by Maiwald [14] (triangles) compared
with simulation results using nonlinearity as a power law with the
exponent = 0.25 for levels above 15 dB.
L
= 20 · lg(p2 ) − 20 · lg(p1 )
dB
= 20 · lg y −1 [y(p) + N ] − 20 · lg y −1 [y(p)]
20 · N d ln{y −1 [y(p)]}
·
ln(10)
dy(p)
20 · N 1
=
· p·dy(p)
ln(10)
≈
(15)
dp
in a double logarithmic diagram we get, using y = p 20 · N L
lg
= lg
(16)
− · lg(p).
dB
ln(10)
Amplitude- as well as frequency-modulated tones generate envelope fluctuations in the individual channels of the
ear-related filter bank. These fluctuations are analyzed with
respect to maxima and minima of the specific loudness values. It is assumed that the fluctuations can be detected if the
difference between the extremes is at least N .5
A comparison of simulation results and measurements for
just-noticeable amplitude differences and amplitude modulations is shown in Fig. 6 respectively in Fig. 7.
We conclude that the nonlinear relation between the specific loudness and sound pressure amplitude can be modeled
as a power law with piecewise-defined exponents : = 1
for levels near hearing threshold, ≈ 0.2–0.25 at medium
levels, and ≈ 0.05–0.15 at high levels.
There is more evidence for a stronger compressive nonlinear behavior of our hearing system (power law with an
exponent ≈ 0.2–0.25) than the one used in ISO 523B,
coming from the level dependence of roughness calculation
5 For simulating the influence of modulation frequency on the justnoticeable amplitude and frequency modulations the time weighting (“the
ear’s temporal window”) is of especial importance. A third-order low-pass
filter according to Eq. (8) has been implemented [1].
Fig. 8. Degree of modulation m of an amplitude-modulated tone
(f = 1 kHz, constant level: L = 80 dB) which produces the same
roughness as a modulated tone of equal frequency with m = 1 and
sound pressure level L. The modulation frequency is fmod =70 Hz.
Data are taken from Terhardt [17]; solid line: simulation using the
roughness calculation based on the described hearing model.
(based on the hearing model) [1] (see Fig. 8 for good correspondence between calculated results and subjective data).6
4. Application of the hearing model to sound
quality evaluation
The Relative Approach method [18] is an analysis tool
developed to model a major characteristic of human hearing. This characteristic is the much stronger subjective response to patterns (tones and/or relatively rapid time-varying
structure) than to slowly changing levels and loudnesses.
6 In general the nonlinear processing is modeled using an even more
compressive nonlinearity followed by a correlation analysis, which for
stationary signals behaves as the above-described power law: for details,
refer to [15,16].
R. Sottek, K. Genuit / Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
163
Hearing model spectrum vs. time
Nonlinear transform
according to the
hearing model
Regression vs. time for
each frequency band
Smoothing operation
vs. frequency
Smoothing operation
vs. frequency
Regression vs. time for
each frequency band
Nonlinear transform
according to the
hearing model
Nonlinear transform
according to the
hearing model
Nonlinear transform
according to the
hearing model
f
f
f−g
g
g
Set subthreshold
values to zero
Relative Approach
analysis for tonal
components
Relative Approach
analysis for
transient signals
f−g
Set subthreshold
values to zero
+
Relative Approach analysisboth for
time and frequency patterns
Fig. 9. Block diagram of the Relative Approach analysis based on the hearing model spectrum versus time.
Fig. 10. Noise of a PC workstation, hard disk access. Relative Approach analysis, optimized for sensitivity to both temporal and
tonal patterns. Time scale is horizontal, frequency vertical; color indicates magnitude. Regarding the terminology of Relative Approach
measurement units: due to the nonlinearity in the relationship between sound pressure and perceived loudness, the term “compressed
pressure” in compressed Pascals (cPa) is used to describe the result of applying the nonlinear transform. Note the time structure in some
of the tones, as well as the fine time structure in the “seek” operations (starting at 3.6 s).
164
R. Sottek, K. Genuit / Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
It is assumed that human hearing creates for its automatic
recognition process a running reference sound (an “anchor
signal”) against which it classifies tonal or temporal pattern information moment-by-moment. It evaluates the difference between the instantaneous pattern in both time and
frequency and the “smooth” or less-structured content in
similar time and frequency ranges. In evaluating the acoustic
quality of a patterned situation, the absolute level or loudness is almost completely without significance. Temporal
structures and spectral patterns are important factors in deciding whether a sound makes an annoying or disturbing
impression.
The Relative Approach has subsequently been expanded
in scope. Various time-dependent spectral analyses can be
used as pre-processing for the Relative Approach; not only
FFT-based analyses but also the hearing model spectrum
versus time.
Recent extensions of the method give the user a choice
of combining time-sensitive and frequency-sensitive procedures, with adjustable priority weighting between the two
and independent settings choices for each. In this way, both
time and frequency patterns in a sound situation may be displayed in the same measurement same measurement result.
A block diagram of the complete Relative Approach analysis is represented in Fig. 9. Fig. 10 shows a Relative Approach 3D analysis of a PC workstation performing a hard
disk access [19] containing a potpourri of tonal and temporal patterns.
The Relative Approach algorithm objectivizes pattern(s)
in accordance with perception by resolving, or extracting,
them while largely rejecting pseudostationary energy. At the
same time, it considers the context of the relative difference
of the “patterned” and “nonpatterned” magnitudes.
5. Conclusion
A theoretical system model of the monaural signal
processing in the human auditory system has been presented which can explain and describe not only special aspects of sound perception, but also a wide range
of psychoacoustic phenomena. The focus has been laid
on a functional scheme to calculate the excitation pattern on the basilar membrane and to model the nonlinear behavior with respect to just-perceptible amplitude
differences and variations. The calculation of elementary psychoacoustic parameters, the simulation of masking thresholds and the just-perceptible frequency differences and variations are other application examples [1].
The deconvolution algorithm, which represents a general mathematical principle, has been applied to the excitation pattern for the detection of dominant spectral
components. This algorithm delivers a high-resolution
spectral analysis with a small product of time and frequency resolution. The application of the hearing model
to sound quality tasks allows evaluating the temporal
and spectral patterns of a sound (“Relative Approach” analysis) where absolute level or loudness is often without
significance.
References
[1] Sottek R. Modelle zur Signalverarbeitung im menschlichen
Gehör. Dissertation, RWTH Aachen, Germany, 1993.
[2] Zwicker E, Feldtkeller R. Das Ohr als Nachrichtenempfänger.
2nd ed., Stuttgart: S. Hirzel Verlag; 1967.
[3] Moore BCJ, Glasberg BR. Suggested formulae for calculating
auditory filter bandwidths and excitation patterns. J Acoust
Soc Am 1983;74(3):750–3.
[4] Moore BCJ, Glasberg BR. The role of frequency selectivity
in the perception of loudness, pitch and time. In: Moore BCJ
(Ed.). Frequency selectivity in hearing. London: Academic
Press; 1986.
[5] Patterson RD, Nimmo-Smith I, Weber FL, Milroy R. The
deterioration of hearing with age: frequency selectivity, the
critical ratio, the audiogram, and speech threshold. J Acoust
Soc Am 1982;72(6):1788–803.
[6] Sottek R, Illgner K, Aach T. An efficient approach to
extrapolation and spectral analysis of discrete signals. In:
ASST ’90 - 7. Aachener Symposium für Signaltheorie. Berlin:
Springer; 1990.
[7] Zwicker E. Über psychologische und methodische
Grundlagen der Lautheit. Acustica 1958;8:237–58.
[8] Stevens SS. The measurement of loudness. J Acoust Soc Am
1955;27(5):815–29.
[9] Hellbrück J. Loudness scaling: ratio scales, rating scales,
intersubject variability and some other problems. In:
Contributions to psychological acoustics. Results of the
fifth Oldenburg symposium on psychological acoustics.
Oldenburg: BIS University of Oldenburg; 1991.
[10] Heller O. Oriented category scaling of loudness
and speechaudiometric validation. In: Contributions to
psychological acoustics. Results of the fifth Oldenburg
symposium on psychological acoustics. Oldenburg: BIS
University of Oldenburg; 1991.
[11] Garner WR. A technique and a scale for loudness
measurement. J Acoust Soc Am 1954;26(2):73–88.
[12] Sellick PM, Patuzzi R, Johnston BM. Measurement of
the basilar motion in the guinea pig using the Mössbauer
technique. J Acoust Soc Am 1982;72(1):131–41.
[13] Schorer E. Vergleich eben erkennbarer Unterschiede und
Variationen der Frequenz und Amplitude von Schallen.
Acustica 1989;68:183–99.
[14] Maiwald D. Die Berechnung von Modulationsschwellen mit
Hilfe eines Funktionsschemas. Acustica 1967;18:193–207.
[15] Sottek R. Gehörgerechte Rauhigkeitsberechnung. Proceedings of the DAGA ’94, Dresden, 1994. p. 1197–200.
[16] Sottek R, Vranken P, Kaiser H-J. Anwendung der
gehörgerechten Rauhigkeitsberechnung. Proceedings of the
DAGA ’94, Dresden, 1994. p. 1201–04.
[17] Terhardt
E.
Über
akustische
Rauhigkeit
und
Schwankungsstärke. Acustica 1968;20:215–24.
[18] Genuit K. Objective evaluation of acoustic quality based
on a relative approach. Proceedings of the inter-noise ’96,
Liverpool, 1996. p. 3233–8.
R. Sottek, K. Genuit / Int. J. Electron. Commun. (AEÜ) 59 (2005) 157 – 165
[19] Bray WR. Using the “Relative Approach” for direct
measurement of patterns in noise situations. Sound and
Vibration, Sep 2004; 20–2.
Klaus Genuit, Dr.-Ing., was born in
Düsseldorf in 1952. He studied Electronic Engineering at the Technical
University of Aachen from 1971 to
1976 and studied economics at the
same university until 1979. At the
same time, he was at the Institute of
“Elektrische Nachrichtentechnik” to
investigate different psychoacoustical
effects of the human hearing. He took
his doctor’s degree in 1984 at the
Technical University of Aachen with his work on “A Model for
Description of the External-Ear-Transfer-Function”. For the next
2 years he was leading the psychoacoustical working group at this
Institute dealing with binaural signal processing, speech intelligibility, hearing aids and telephone systems. In cooperation with
Daimler Benz (Stuttgart), he developed a new, improved artificial
head measurement system for the diagnosis and analysis of sound,
with attributes which are comparable to those of the human ear.
He founded the company “HEAD acoustics GmbH” in 1986 which
is working in the area of binaural signal processing, particularly
concentrating on human hearing equivalent sound-field analysis,
auralization of virtual environment, NVH analysis and measurements in the field of telecommunication.
HEAD acoustics now has 100 employees. It works also in areas
such as telecommunication technology and the virtual simulation
of hearing events. Since 1980 he has applied for several patents
concerning binaural subjects. Dr. Genuit is also the author of nu-
165
merous papers published in connection with various national and
international congresses. He is a member of various associations,
such as AES, ASA, JAS, DAL, SAE, DEGA, NTG, VDI, Additionally he participates in several working groups dealing with the
standardization of measurement regulations as in ITU (formerly
CCITT), DIN, DKE, NALS.
Roland Sottek, Dr.-Ing. was born in
Erftstadt in 1961. He studied Electronic
Engineering at the Technical University
of Aachen, where he received a diploma
in 1987. He took his doctor’s degree in
1993 with his dissertation on a “Signal
Processing Model of the Human Auditory System”. From 1987 to 1988 he
worked at the Philips Research Laboratory Aachen. In 1989 he joined HEAD
acoustics where he started working as
a scientific consultant. From 1999 to 2002 Dr. Sottek was leading
the HEAD consult NVH department. Since October 2002 he has
been in charge of the newly established HEAD research NVH
department. Dr. Sottek has applied for patents concerning the
design of permanent magnet excited motors and signal analysis
techniques. He is author or co-author of about 55 papers published
on the occasion of various national and international conferences
and congresses. Current research work is dealing with models of
human hearing, psychoacoustics, noise engineering, digital signal
processing as well as experimental and numerical methods for
sound-field calculation. He is working in the area of human hearing
equivalent sound-field analysis, auralization of virtual environment
and NVH analysis.