Preface for Paperback Edition

In 1968, Part I of Detection, Estimation, and Modulation Theory [VT68] was pub-
lished. It turned out to be a reasonably successfid book that has been widely used by
several generations of engineers. There were thirty printings, but the last printing
was in 1996. Volumes I1 and 111 ([VT7 1a], [VT7 1b]) were published in 1971 and fo-
cused on specific application areas such as analog modulation, Gaussian signals
and noise, and the radar-sonar problem. Volume I1 had a short life span due to the
shift from analog modulation to digital modulation. Volume 111 is still widely used
as a reference and as a supplementary text. In a moment of youthful optimism, I in-
dicated in the the Preface to Volume 111 and in Chapter 111-14 that a short mono-
graph on optimum array processing would be published in 1971. The bibliography
lists it as a reference, Optimum Array Processing, Wiley, 1971, which has been sub-
sequently cited by several authors. After a 30-year delay, Optimum Array Process-
ing, Part IV of Detection, Estimation, and Modulation Theory will be published this
year.
A few comments on my career may help explain the long delay. In 1972, MIT
loaned me to the Defense Communication Agency in Washington, D.C. where I
spent three years as the Chief Scientist and the Associate Director ofTechnology. At
the end of the tour, I decided, for personal reasons, to stay in the Washington, D.C.
area. I spent three years as an Assistant Vice-president at COMSAT where my
group did the advanced planning for the INTELSAT satellites. In 1978, I became
the Chief Scientist of the United States Air Force. In 1979, Dr. Gerald Dinneen, the
former Director of Lincoln Laboratories, was serving as Assistant Secretary of De-
fense for C3I. He asked me to become his Principal Deputy and I spent two years in
that position. In 1981, I joined M/A-COM Linkabit. Linkabit is the company that Ir-
win Jacobs and Andrew Viterbi had started in 1969 and sold to M/A-COM in 1979.
I started an Eastern operation which grew to about 200 people in three years. After
Irwin and Andy left M/A-COM and started Qualcomm, I was responsible for the
government operations in San Diego as well as Washington, D.C. In 1988, M/A-
COM sold the division. At that point I decided to return to the academic world.
I joined George Mason University in September of 1988. One of my priorities
was to finish the book on optimum array processing. However, I found that I needed
to build up a research center in order to attract young research-oriented faculty and

vii
viii Prefacefor Paperback Edition

doctoral students. The process took about six years. The Center for Excellence in
Command, Control, Communications, and Intelligence has been very successful
and has generated over $300 million in research funding during its existence. Dur-
ing this growth period, I spent some time on array processing but a concentrated ef-
fort was not possible. In 1995, I started a serious effort to write the Array Process-
ing book.
Throughout the Optimum Array Processing text there are references to Parts I
and I11 of Detection, Estimation, and Modulation Theory.The referenced material is
available in several other books, but I am most familiar with my own work. Wiley
agreed to publish Part I and 111 in paperback so the material will be readily avail-
able. In addition to providing background for Part IV, Part I is still useful as a text
for a graduate course in Detection and Estimation Theory. Part I11 is suitable for a
second level graduate course dealing with more specialized topics.
In the 30-year period, there has been a dramatic change in the signal processing
area. Advances in computational capability have allowed the implementation of
complex algorithms that were only of theoretical interest in the past. In many appli-
cations, algorithms can be implemented that reach the theoretical bounds.
The advances in computational capability have also changed how the material is
taught. In Parts I and 111, there is an emphasis on compact analytical solutions to
problems. In Part IV, there is a much greater emphasis on efficient iterative solu-
tions and simulations. All of the material in parts I and I11 is still relevant. The books
use continuous time processes but the transition to discrete time processes is
straightforward. Integrals that were difficult to do analytically can be done easily in
[email protected] various detection and estimation algorithms can be simulated and
their performance compared to the theoretical bounds. We still use most of the prob-
lems in the text but supplement them with problems that require Matlab@solutions.
We hope that a new generation of students and readers find these reprinted edi-
tions to be useful.

HARRY
L. VANTREES
Fairjkx, Virginia
June 2001
 Preface

The area of detection and estimation theory that we shall study in this book
represents a combination of the classical techniques of statistical inference
and the random process characterization of communication, radar, sonar,
and other modern data processing systems. The two major areas of statis-
tical inference are decision theory and estimation theory. In the first case
we observe an output that has a random character and decide which of two
possible causes produced it. This type of problem was studied in the middle
of the eighteenth century by Thomas Bayes [l]. In the estimation theory
case the output is related to the value of some parameter of interest, and
we try to estimate the value of this parameter. Work in this area was
published by Legendre [2] and Gauss [3] in the early nineteenth century.
Significant contributions to the classical theory that we use as background
were developed by Fisher [4] and Neyman and Pearson [5] more than
30 years ago. In 1941 and 1942 Kolmogoroff [6] and Wiener [7] applied
statistical techniques to the solution of the optimum linear filtering
problem. Since that time the application of statistical techniques to the
synthesis and analysis of all types of systems has grown rapidly. The
application of these techniques and the resulting implications are the
subject of this book.
This book and the subsequent volume, Detection, Estimation, and
Modulation Theory, Part 11, are based on notes prepared for a course
entitled “Detection, Estimation, and Modulation Theory,” which is taught
as a second-level graduate course at M.I.T. My original interest in the
material grew out of my research activities in the area of analog modulation
theory. A preliminary version of the material that deals with modulation
theory was used as a text for a summer course presented at M.I.T. in 1964.
It turned out that our viewpoint on modulation theory could best be
understood by an audience with a clear understanding of modern detection
and estimation theory. At that time there was no suitable text available to
cover the material of interest and emphasize the points that I felt were
ix
x Preface

important, so I started writing notes. It was clear that in order to present

the material to graduate students in a reasonable amount of time it would
be necessary to develop a unified presentation of the three topics: detection,
estimation, and modulation theory, and exploit the fundamental ideas that
connected them. As the development proceeded, it grew in size until the
material that was originally intended to be background for modulation
theory occupies the entire contents of this book. The original material on
modulation theory starts at the beginning of the second book. Collectively,
the two books provide a unified coverage of the three topics and their
application to many important physical problems.
For the last three years I have presented successively revised versions of
the material in my course. The audience consists typically of 40 to 50
students who have completed a graduate course in random processes which
covered most of the material in Davenport and Root [8]. In general, they
have a good understanding of random process theory and a fair amount of
practice with the routine manipulation required to solve problems. In
addition, many of them are interested in doing research in this general area
or closely related areas. This interest provides a great deal of motivation
which I exploit by requiring them to develop many of the important ideas
as problems. It is for this audience that the book is primarily intended. The
appendix contains a detailed outline of the course.
On the other hand, many practicing engineers deal with systems that
have been or should have been designed and analyzed with the techniques
developed in this book. I have attempted to make the book useful to them.
An earlier version was used successfully as a text for an in-plant course for
graduate engineers.
From the standpoint of specific background little advanced material is
required. A knowledge of elementary probability theory and second
moment characterization of random processes is assumed. Some familiarity
with matrix theory and linear algebra is helpful but certainly not necessa.ry.
The level of mathematical rigor is low, although in most sections the results
could be rigorously proved by simply being more careful in our derivations.
We have adopted this approach in order not to obscure the important
ideas with a lot of detail and to make the material readable for the kind of
engineering audience that will find it useful. Fortunately, in almost all
cases we can verify that our answers are intuitively logical. It is worthwhile
to observe that this ability to check our answers intuitively would be
necessary even if our derivations were rigorous, because our ultimate
objective is to obtain an answer that corresponds to some physical system
of interest. It is easy to find physical problems in which a plausible mathe-
matical model and correct mathematics lead to an unrealistic answer for the
original problem.
 We have several idiosyncrasies that it might be appropriate to mention.
In general, we look at a problem in a fair amount of detail. Many times we
look at the same problem in several different ways in order to gain a better
understanding of the meaning of the result. Teaching students a number of
ways of doing things helps them to be more flexible in their approach to
new problems. A second feature is the necessity for the reader to solve
problems to understand the material fully. Throughout the course and the
book we emphasize the development of an ability to work problems. At the
end of each chapter are problems that range from routine manipulations to
significant extensions of the material in the text. In many cases they are
equivalent to journal articles currently being published. Only by working a
fair number of them is it possible to appreciate the significance and
generality of the results. Solutions for an individual problem will be
supplied on request, and a book containing solutions to about one third
of the problems is available to faculty members teaching the course. We
are continually generating new problems in conjunction with the course
and will send them to anyone who is using the book as a course text. A
third issue is the abundance of block diagrams, outlines, and pictures. The
diagrams are included because most engineers (including myself) are more
at home with these items than with the corresponding equations.
One problem always encountered is the amount of notation needed to
cover the large range of subjects. We have tried to choose the notation in a
logical manner and to make it mnemonic. All the notation is summarized
in the glossary at the end of the book. We have tried to make our list of
references as complete as possible and to acknowledge any ideas due to
other people.
A number of people have contributed in many ways and it is a pleasure
to acknowledge them. Professors W. B. Davenport and W. M. Siebert have
provided continual encouragement and technical comments on the various
chapters. Professors Estil Hoversten and Donald Snyder of the M.I.T.
faculty and Lewis Collins, Arthur Baggeroer, and Michael Austin, three
of my doctoral students, have carefully read and criticized the various
chapters. Their suggestions have improved the manuscript appreciably. In
addition, Baggeroer and Collins contributed a number of the problems in
the various chapters and Baggeroer did the programmingnecessary for many
of the graphical results. Lt. David Wright read and criticized Chapter 2.
L. A. Frasco and H. D. Goldfein, two of my teaching assistants, worked
all of the problems in the book. Dr. Howard Yudkin of Lincoln Laboratory
read the entire manuscript and offered a number of important criticisms.
In addition, various graduate students taking the course have made
suggestions which have been incorporated. Most of the final draft was
typed by Miss Aina Sils. Her patience with the innumerable changes is
xii Preface

sincerely appreciated. Several other secretaries, including Mrs. Jarmila

Hrbek, Mrs. Joan Bauer, and Miss Camille Tortorici, typed sections of the
various drafts.
As pointed out earlier, the books are an outgrowth of my research
interests. This research is a continuing effort, and I shall be glad to send our
current work to people working in this area on a regular reciprocal basis.
My early work in modulation theory was supported by Lincoln Laboratory
as a summer employee and consultant in groups directed by Dr. Herbert
Sherman and Dr. Barney Reiffen. My research at M.I.T. was partly
supported by the Joint Services and the National Aeronautics and Space
Administration under the auspices of the Research Laboratory of Elec-
tronics. This support is gratefully acknowledged.
Harry L. Van Trees
Cambridge, Massachusetts
October, 1967.

REF ERE NCE S

Thomas Bayes, “An Essay Towards Solving a Problem in the Doctrine of

Chances,” Phil. Trans, 53, 370-418 (1764).
A. M. Legendre, Nouvelles Mithodes pour La Ditermination ces Orbites des
ComPtes, Paris, 1806.
K. F. Gauss, Theory of Motion of the Heavenly Bodies Moving About the Sun in
Conic Sections, reprinted by Dover, New York, 1963.
R. A. Fisher, “Theory of Statistical Estimation,” Proc. Cambridge Philos. SOC.,
22, 700 (1925).
J. Neyman and E. S. Pearson, “On the Problem of the Most Efficient Tests of
Statistical Hvuotheses.” Phil. Trans. Rov. SOC.London. A 231. 289. (1933).
A. Kolmogdioff, “Inierpolation and Extrapolation von Stadonlren Zufalligen
Folgen,” Bull. Acad. Sci. USSR,Ser. Math. 5, 1941.
N. Wiener, Extrapolation, Interpolation, and Smoothing of Stationary Time Series,
Tech. Press of M.I.T. and Wiley, New York, 1949 (originally published as a
classified report in 1942).
W. B. Davenport and W. L. Root, Random Signals and Noise, McGraw-Hill,
New York, 1958.