Performance characterization of matched subspace detector for spectrum sensing

Abstract

Matched Subspace Detector (MSD) is a robust detection scheme used for detection of the target primary user signal buried in high-dimensional noise where the target signal is assumed to be placed in low-rank subspace. In this paper we attempt to present the benefits of MSD detector by providing the performance comparison with some other existing blind signal detection techniques and further confirmed detector performance on varying signal dimension and false alarm probabilities. For the scenario when the subspace estimation is done from finite, noisy, signal-bearing training data we propose to use information theoretic criteria (ITC) which highlights the importance of using a critical number of informative components which depends on training phase SNR, system dimension and number of training samples.