Preface
As a discipline, signal detection has evolved significantly over the last 40 years.
Some changes have been caused by technical advances, like the development
of robust detection methods, or the use of the theory of large deviations to
characterize the asymptotic performance of tests, but most changes have been
caused by transformations in the engineering systems to which detection tech-
niques are applied. While early applications of signal detection focused on
radar and sonar signal processing or the design of digital communication re-
ceivers, newer areas of application include image analysis and interpretation,
document authentification, biometrics, and sensor or actuator failure detec-
tion. This expanded scope of application has required some adjustment in
standard ways of formulating detection problems. For example, image process-
ing applications typically combine parameter estimation and detection tasks,
so the separation of parameter estimation and detection in distinct operations
typical of early communication systems, where parameter estimation was ac-
complished through the use of training signals, needs to be abandoned. Other
changes have occured in the design of communication systems which make
it increasingly difficult to treat the detection of communications signals and
of radar/sonar signals in a unified manner. This common framework assumes
implicitly that intersymbol interference is not present and that channel cod-
ing and modulation are implemented separately, since in this case modulated
signals can be detected one symbol at a time. But modern communication
systems are typically designed to operate over bandlimited channels where in-
tersymbol interference is present, and starting with the introduction of trellis
coded modulation, modulation and coding have become intertwined. In this
context, the detection of modulated signals can no longer be treated on a
symbol-by-symbol basis but needs to be viewed as a sequence detection prob-
lem, where the sequence is generated by a Markov chain. Another feature of
modern radar and communication systems, in particular wireless systems, is
that they often need to operate in a rapidly changing environment. So even
if training or calibration signals are available to estimate the system param-
eters, because parameters may change quickly, it is desirable to constantly
