Estimating Randomly Modulated Periodic Processes Generated from Nonlinear Mechanisms
Speaker: Prof Melvin Hinich, University of Texas at Austin
When: 2005-08-23 16:00:00
Venue: 78-621 (venue may change)
Host: Dr. Ariel Liebman
Abstract:Applied Research Laboratories of the University of Texas at Austin
There is always variation in the amplitude and phase of what is
usually called a periodic signal. The periodicity is really in the
expected value of the process from year to year. The variation is
due to random effects that disturb the basic periodicity. This
"wobbly" periodic signal is called a randomly modulated
periodicity. Since a linear system can not generate a stable
periodic output the modulation pattern can be used to identify the
nature of the nonlinear data generating mechanism. In some
applications the modulation obscures the periodic pattern and a
statistical test is required to detect the underlying periodicity. I
define the concept of a signal coherence function of a modulated
periodic signal and apply it to a variety of practical signal
processing problem areas. It is shown that the signal coherence
function can improve the detectability of this type of signal as
compared with the standard periodogram (spectrogram) methods. An
artificial data set was created to compare standard spectrogram
analysis with the signal coherence detector. The average pulse is
reconstructed using the signal coherence function. The FFT of the
mean frame for the data is computed, where the frame length is a
multiple of the pulse length. The amplitudes of the mean frame whose
signal coherence is below a set threshold are zeroed out. This
complex vector is then transformed into the time domain by an FFT
operation. This reconstructed signal is called the coherent part of
the signal.
Biography:Professor Melvin Hinich principal research areas are in the fields
of Analytical Political Science, Time Series Analysis, Economics,
and Statistical Theory and Methods in Engineering and Science with
particular emphasis on Signal Processing. He has published widely in
leading journals in these fields as the following small sample
indicates: Journal of the American Statistical Association, Journal
of Time Series Analysis, Annals of Mathematical Statistics, Journal
of Econometrics, Econometrica, Journal of Economic Theory, Journal
of the Acoustical Society of America, Technometrics, IEEE
Transactions on Signal Processing and Signal Processing. Professor
Hinich received his Ph.D in Statistics from Stanford University in
1963 and is currently Professor of Government and Economics at the
University of Texas at Austin and is also a Research Scientist in
the Applied Research Laboratories of the University of Texas. He has
held academic positions previously at Virginia Tech (Economics) and
Carnegie-Mellon University (Industrial Administration and
Statistics). He has also done consultancy work for Bell
Laboratories, Columbia University's Hudson Laboratories, the
U.S. Navy, and a number of other organizations. He is a Fellow of
the American Statistical Association, the Institute of Mathematical
Statistics, and the Public Choice Society.
Type: ACCS
Contact:Dr. Ariel Liebman, seminar host (aliebman@itee.uq.edu.au)
or Guido Governatori (ITEE seminar co-ordinator)
(guido@itee.uq.edu.au)