Polarization-Based Spectrum Sensing Algorithms for Cognitive Radios: Upper and Practical Bounds and Experimental Assessment

Abstract

Spectrum sensing exploiting polarization has been identified as an effective method for improving detection performance. However, the optimal polarization-based sensing algorithm with an upper bound performance has yet to be developed. The polarization likelihood ratio test (PLRT) is derived using the Neyman–Pearson theorem, resulting in an optimal detector with regard to the likelihood ratio. The best linear polarization processor or filter, i.e., the polarization matched filter (PMF), which processes the observed polarization data to provide the maximum signal-to-noise ratio (SNR) to the output of the filter, is also derived. Although the PLRT and the PMF provide an ideal upper bound on the performance of all detection methods that exploit polarization when likelihood functions or statistic parameters of likelihood functions are exactly known, their performance for realistic blind sensing remains unknown. Using the generalized likelihood ratio test (GLRT) method, we formulate two practical polarization-based detectors, i.e., GLRT based on received electric-field vector (GLRT-EV) and GLRT of the Stokes subvector, according to the statistical distribution of observed data expressed as a received electric-field vector and a Stokes vector, respectively. The two GLRT detectors can serve as a practical bound when all statistical parameters are unknown. Theoretical and numerical simulation results of the detection performance for the proposed detectors are presented. Additionally, these polarization-based detectors were thoroughly tested via a realistic wireless regional area network. Our experimental results show that the GLRT-EV detector exhibits better performance than other polarization-based detectors in terms of sample complexity and computational complexity, which is particularly critical in real-time applications.