Multitaper spectrum based detection for spectrum sensing in cognitive radio networks

Abstract

Given the ever-increasing demand for high data-rate transmission, efficient use of the precious electromagnetic spectrum poses a challenging issue to the wireless communications community. One possible solution is the use of cognitive radio (CR), which was incepted by Mitola in 1999 and has shortly become a frontier research focus in the past few years. CR networks offer improved service quality, allowing users to operate in a certain allocated frequency band without a license and without interfering with the primary user of the very band. Spectrum sensing, a key enabling function of CR networks, needs to reliably detect primary user signals of possibly unknown types in relatively low SNR. Reliability and rapid reaction, which implies low complexity in computation, are two basic requirements for a sensing algorithm adopted in CR networks. Fully exploiting the magnitude and phase information in received data is crucial to the detection of cognitive radio signals, and this is usually done in the time domain. In fact, the same set of information can also be exploited in the frequency domain in the form of power spectrum. The most appealing one is the multitaper spectrum (MTS) estimator, which enjoys a high estimation accuracy and relatively low computational complexity making it an attractive candidate for use in CR networks. In our research, we investigate the usage of MTS-based CR detector, aiming to fully understand its capability in exploiting the advantage of MTS. Starting by analyzing the advantages of MTS and discussing its possible application in CR networks, we consider the spectrum sensing problem based on MTS both in single-user scenario and multiple-user scenario. For the single-user scenario, the decision variable of MTS-based detector can be represented as a quadratic function of Gaussian vector, and based on this fact we evaluate its detection performance in the framework of Neyman-Pearson Criterion. We obtain simple expressions for detection probability and constant false-alarm rate whereby a threshold can be easily determined. For the multiple-user scenario, we propose a centralized cooperative sensing scheme in which an eigen-based data fusion rule is adopted at the fusion center which collects MTS-based estimates of cognitive users. The proposed fusion rule is advantageous in that it fully exploits the inherent correlation structure among data from different users, so that a good overall detection performance is obtained. The superiority of MTS-based detectors over commonly adopted techniques is theoretically analyzed and confirmed through simulation results for both scenarios.