11.1 INTRODUCTION

In signal detection, the basic goal is to derive the optimum receiver structure based on some criterion that is determined by the application. Such optimal (ideal) detectors require an essentially complete statistical description of the input signals and noise. In practice, this information may not be available a priori, and the statistics of the input data may also vary with time. These constraints lead to the consideration of other (nonoptimal) detectors.

In practical radar signal detection systems, the problem is to automatically detect a target in thermal noise plus clutter. Clutter is the term applied to any unwanted radar signal from scatterers that are not of interest to the radar user ^[1]. Examples of unwanted echoes, or clutter, in radar signal detection are reflections from terrain, sea, rain, birds, insects, chaff, and so on. Chaff consisting of dipole reflectors, usually metallic, is dropped from aircrafts to mask the real target from the radar. Due to the development of Doppler processing, it is now mainly used for slow moving targets ^[2]. Since the environment in which a radar operates depends on factors such as weather conditions and the physical location of operation, the returned signals are statistically nonstationary, with unknown variance at the receiver input. Thus, the ideal detector using a fixed threshold is extremely sensitive to the total noise (thermal noise plus clutter) variance. In fact, a small increase in the total noise power results in a corresponding increase of several orders of magnitude...