Estimation-based noise-robust sensing

Abstract

It is well known that noise-modeling uncertainties give rise to a fundamental limit on the sensitivity of energy-based detectors (or related moment-based detectors), also known as the “SNR wall”. Similar “walls” appear also when localizing unknown sources for various applications (including cognitive radio), since localization techniques are based on tempo-spatial signal features. Low SNR regimes plus vague knowledge of the characteristics of the signal under detection in such applications motivate the exploration of techniques that are robust to such parameter uncertainties and therefore immune to such limits. A simple conceptual framework is proposed herein that naturally generates techniques with such immunity advantages for detecting the presence of an unknown source in broadband noise. At the cost of longer observation time, they allow SNR-wall-free detection. They also lend themselves to precise analysis and parametric optimization.