Modeling auditory processing of amplitude modulation
by Torsten Dau
This document contains the three abstracts of the submitted
Doctoral Thesis at the University of Oldenburg, Fachbereich Physik.
Modulation detection and masking with narrowband carriers (Abstract)
The chapter presents a quantitative model for describing modulation
detection and modulation masking experiments. The model combines
several stages of preprocessing with a decision device that has the
properties of an optimal detector. In order to analyze the temporal
envelope of the signals, the model comprises a modulation filterbank.
The filterbank exhibits two domains with different scaling. In the
range 0-10 Hz, the filters have a constant bandwidth of 5 Hz. Between
10 Hz and 1000 Hz a logarithmic scaling with a constant Q-value of 2
was assumed.
To preclude spectral effects in temporal processing, measurements and
corresponding simulations were performed with stochastic
narrowband noise carriers at a high center frequency (5 kHz). The
bandwidth of the modulated signals was chosen in order to be smaller
than the bandwidth of the stimulated peripheral filter. Model
predictions were compared with the performance of human
observers. Simulated and psychophysically determined thresholds were
estimated with a 3-Interval Forced-Choice adaptive procedure.
For conditions in which the modulation frequency was smaller than half
the bandwidth of the carrier, the model accounts for the lowpass
characteristic in the threshold functions, which is well known from
other measurements of the temporal modulation transfer function (TMTF)
in the literature [e.g. Viemeister, JASA 66, 1364--1380 (1979)]. This
lowpass characteristic is predicted by the model because of the
logarithmic scaling of the modulation filter bandwidth, leading to a
3-dB increase in modulation detection threshold per doubling of the
modulation frequency. In conditions where the modulation frequency is
larger than half the bandwidth of the carrier, the model can account
for the highpass characteristic in the threshold function, reflecting
the auditory system's modulation frequency selectivity.
In a further experiment, a classical masking paradigm for
investigating frequency selectivity was adopted and translated to the
modulation frequency domain. A sinusoidal carrier at 5 kHz was
used. The masker modulation consisted of the third through seventh
component of a tone complex with a fundamental of 30 Hz. The subject's
task was to detect a sinusoidal test modulation in the presence of the
competing tone-complex masker. The results provide further evidence of
modulation-frequency-specific channels. In all cases, the model
describes the experimental data to within a few dB.
Spectral and temporal integration in amplitude modulation detection (Abstract)
A multi-channel model is proposed to describe effects of spectral and
temporal integration in amplitude modulation detection for a
stochastic noise carrier and sinusoidal amplitude modulation. The
model is based on the modulation filterbank concept which was
established in the previous chapter for modulation perception in
narrowband conditions (single-channel model). To integrate information
across frequency, the detection process of the model optimally weights
and linearly combines the channel outputs, assuming independent
observations at the outputs. To integrate information across time, a
``multiple-look'' strategy is realized within the detection stage of
the model.
The model accounts for the findings by Eddins [JASA 93, 470--479
(1993)] that, a), the ``time constants'' associated with the temporal
modulation transfer functions (TMTF) derived for narrowband stimuli do
not vary with carrier frequency region and, b), the time constants
associated with the TMTFs decrease monotonically with increasing
stimulus bandwidth. The model is able to predict masking patterns in
the modulation-frequency domain, as they were observed experimentally
by Houtgast [JASA 85, 1676--1680 (1989)]. The model also accounts for
the finding by Sheft and Yost [JASA 88, 796--805 (1990)] that the long
``effective'' integration time constants derived from the data are two
orders of magnitude larger than the time constants derived from the
cut-off frequency of the TMTF.
The combination of the modulation filterbank concept and the optimal
decision algorithm, proposed here, appears to present a powerful
strategy for describing modulation detection phenomena in narrowband
and broadband conditions.
Amplitude modulation detection with sinusoidal carriers at different
frequencies (Abstract)
For a set of carrier frequencies (2, 3, 4, 5, 9 kHz), the temporal
modulation transfer function (TMTF) was measured in the range of
modulation frequencies from 10-800 Hz. These conditions include the
range of modulation frequencies in which the cue for modulation
detection changes from a temporal (at low modulation rates) to a
spectral (at higher rates) cue. The data of all subjects show a flat
threshold function up to 100 Hz, independent of the carrier
frequency. For modulation rates larger than 100 Hz, detection
thresholds increase at a rate of about 2-5 dB/oct., depending on
center frequency, and finally decrease when the detection cue changes
to detection of the spectral sidebands of the modulation. The flat
threshold function up to a modulation of about 100 Hz contradicts the
hypothesis of a modulation lowpass filter at about 50-60 Hz as often
discussed in the literature.
Results from simulations obtained with the modulation filterbank model were
compared with the data. The model which has successfully been
applied to modulation detection conditions with narrowband and
broadband noise carriers, accounts for the flat threshold up to a
modulation rate of 100 Hz and also accounts for the
frequency-dependent roll-off in the threshold function.
However, the model does not account for the increase of experimental
thresholds for modulation rates above 100 Hz (for high center frequencies).
A certain loss of information seems to occur in the ``central'' auditory
processing of high modulation rates which is not included in the
present model. Further experiments and modeling efforts are required
to clarify the processing of fast modulations (> 100 Hz) in the
auditory system.
torsten@medi.physik.uni-oldenburg.de
Last modified:
Tue Mar 19 12:14:55 1996