A geometric and multiresolution analysis approach to robust detection

Abstract

Summary form only given, as follows. Detection algorithms, whose design takes into account prior knowledge about the signals and the channel, face a quandary: they provide marked improvement in performance when the field operating conditions match well this available knowledge; but they experience strong degradation when the actual conditions depart from the assumed ones. In other words, high resolution and robustness are commonly at odds. A third important variable affecting this tradeoff is the computational complexity of the solution. A geometric based approach to designing detectors leads to a satisfying compromise: simple to implement detectors that are robust to mismatches and that exhibit good performance. The approach designs a representation subspace that is a good approximation (in the gap metric sense) to the signal set (a priori information), and uses multiresolution and wavelet analysis to design the representation subspace and implement the detector. The approach can be applied to multipath channels, and detection results illustrate the robustness of the geometric gap detector.