Using NLMS Algorithms in Cyclostationary-Based Spectrum Sensing for Cognitive Radio Networks

Abstract

In general, signals transmitted by primary users (PUs) in a cognitive radio network have cyclostationary characteristics, whereas, the noise signals do not have any cyclostationary characteristics. Thus, detecting the existence of the PUs can be performed by measuring the cyclostationarity of the signals which are present in the communication channels. In this paper, we propose a sensing algorithm for secondary users (SUs) that uses a set of normalized least mean square (NMLS) adaptive filters in order to estimate the signal in the communication channel from its frequency shifted samples. When the received signal is cyclostationary, i.e. the PUs are transmitting, the norm of the NMLS filters' weights at the related SU is anticipated to be nonzero. On the other hand, when the signal is totally the noise, the norm converges to zero. Therefore, in the proposed algorithm, the sensing is made by comparing the norm of weights to a threshold. We derive the probability of detection and false alarm and by simulations, we compare the performance of our algorithm to other known sensing algorithms with respect to the detection probability and complexity.