Orthogonal-mixing adaptive detection in incompletely specified noise

A new class of adaptive detection procedures is proposed based on the principle of orthogonal mixing. The principle is formulated as a theorem and applied to typical detectors in the proposed class. The new detection procedure is implemented in the form of a parallel nonlinear filtering structure which can be digitally realized. In incompletely specified and slowly-varying noise environments, orthogonal-mixing adaptive detectors are shown to be robust and to perform better, from the efficiency point of view, when compared to the robust Huber-Martin-Schwartz (HMS) detectors.