 
Summary: Development of a statistical procedure for detecting
the number of signals in a radar measurement
P.Chen, M.C.Wicks and R.S.Adve
Abstract: Ranking and selection theory is applied to the eigenvalue problem. Of concern is the
development of a procedure for computing the number of signals in a measurement data vector. In
the authors' approach, the multiplicity of the noise eigenvalue is computed, and used in
calculating the number of nonnoise (signal) eigenvalues.
1 Introduction
In the analysis of measured data, an approach that is often
used involves modelling observations as the superposition
of a finite number of signals embedded in additive Gaus
sian noise. This is especially true in phased array signal
processing, timeharmonic analysis, computing the natural
response of a system by estimating the number of poles
from measurement data, and in detccting overlapping
target echocs from radar backscatter. Practical spacetime
adaptivc processing for airborne radar requires effcctive
utilisation of available degrees of frcedom. The question
therefore ariscs how many degrees of frccdom are required
in a given interference scenario. A fundamental issue in
