CYCLIC FEATURE DETECTION USING COMPRESSED SENSING FOR WIDEBAND COGNITVE RADIO

Abstract

Recent surveys on spectrum utilization have shown that fixed allocation scheme is inefficient in utilizing available spectrum. Cognitive radio technology maximizes the utilization of the available spectrum by allowing unlicensed users to transmit in a bandwidth allocated to licensed user whenever the bandwidth is not in use. In order not to interfere with licensed user, cognitive radio must be able to detect the presence of primary users' signal, which may be immersed in noise. Hence it must adhere to strict benchmarks in the quality of spectrum sensing. Several spectrum sensing schemes such as Energy detector, Cyclostationary detector, Matched filter detection etc., are proposed for this purpose. Energy detection is the simplest spectrum sensing scheme. However, an uncertainty in the variance of noise deteriorates the performance of the Energy detector. Cyclostationary detection can exploit the inherent redundancy present in the many human-made signals in the form of spectral correlation. Cyclostationary detector is robust to noise uncertainty and can differentiate signal from interference and noise. Due to the advances in RF and VLSI technology, it is possible to build a trans-receiver which can perform complex tasks such as detecting the presence of multiple primary users simultaneously in a wide band. This may require cognitive radio to perform sensing task at a very high rate. Using compressed sensing technique, the rate at which digital signal processing is to be carried out can be reduced, by exploiting the sparseness in the spectrum of received signal. A compressive framework for wideband spectrum sensing is developed in the literature which exploits cyclostationary properties of the signals transmitted by the users in the band of interest for detecting the vacant frequencies available for transmission and the sparse nature of cyclic spectral density of the received signal to reduce the sampling rate. In compressed sensing framework, sparse signal recovery requires solving a minimization problem. This optimization problem can be solved using Kalman filter with a constraint on norm of the state vector (CSKF) and hence, recovering the sparse signal form a series of noisy observations. The CSKF performs the filtering operation while recovering the sparse signal from compressed samples. Hence, it is intuitive to apply the CSKF algorithm for recovering cyclic spectral density in the compressive framework for wideband spectrum sensing.