Array processing using robust partition statistics

In this paper, the theory of m-interval polynomial approximation (MIPA) is modified and extended to include sequential operation for detecting stochastic weak signals by an array of sensors. The main concern is to formulate the descriptive structure of a class of robust array processing detectors when the functional form of the underlying noise distribution is poorly specified. In particular, we partition the observation space of each sensor into a finite number of regions called intervals based on knowledge of only the quantiles of the noise distribution. The general structure of the robust array consists of two modes of operation, parametric and distribution free, which are switched over depending on the amplitude of the data at each sensor. Next, some truncated and curved boundary decision rules for sequential operation of the detector are introduced. This leads naturally to an efficient operation of the detector even in extremely low signal-to-noise (SNR) environments by eliminating the influence of occasionally unbounded sample sequence that are an integral part of sequential detectors operating in severe noise. The new detectors perform very well when compared with robust array detectors proposed by others