3758 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 8, OCTOBER 2011 Blind Spectrum Sensing Using Antenna Arrays and Path Correlation Mahdi Orooji, Student Member, IEEE, Reza Soosahabi, Student Member, IEEE, and Mort Naraghi-Pour, Member, IEEE Abstract—In this paper, we consider the problem of spectrum sensing in cognitive radios (CRs) when the receiver of the sec- ondary user (SU) is equipped with a multiantenna system. Using an estimate of the cross correlation among the signals received at different antenna elements, we propose a blind detection method, which assumes no prior knowledge of the signaling scheme used by the PU, the noise power, or the channel path coefficients. The cross correlation among the received signals is a result of the correlation among the channel path coefficients from the primary user (PU) transmitter to different antenna elements of the sec- ondary receiver. The detection and false alarm probabilities of the proposed algorithms are evaluated using an asymptotic analysis, and the results are compared with simulation results. It is shown that the proposed methods outperform several recently proposed blind-sensing techniques for CRs using multiple antennas. Index Terms—Blind spectrum sensing, cognitive radio (CR), multiantenna system, opportunistic spectrum access (OSA), path correlation. I. I NTRODUCTION O VERCROWDING of the radio spectrum, coupled with the ever-increasing demand for wireless services, has led to the introduction of a new paradigm in spectrum management, i.e., opportunistic spectrum access (OSA). In OSA, a secondary (unlicensed) user (SU) can identify and utilize the portions of the licensed spectrum that are currently unused by the primary (licensed) users (PU). Cognitive radio (CR), which was first introduced in [1], has been proposed as the enabling technology for OSA. In a CR network, an SU must reliably sense the channel to determine whether another user’s signal (primary or secondary) is present. In the absence of another signal, the SU adapts its operating parameters (e.g., carrier frequency, transmit power, modulation, coding, etc.) to make best use of the avail- able spectrum hole and provide the desired quality of service. Spectrum sensing is therefore a critical function of a CR. Reliable spectrum sensing is a challenging task. A high probability of detection is required to ensure that secondary users do not cause undue interference to PUs who have priority of using the channel. On the other hand, a low false alarm probability is required to increase channel utilization when the Manuscript received February 18, 2011; revised June 7, 2011; accepted July 22, 2011. Date of publication August 30, 2011; date of current version October 20, 2011. This paper was presented in part at the IEEE Global Communications Conference, Houston, TX. The review of this paper was coordinated by Y.-C. Liang. The authors are with the Department of Electrical and Computer Engi- neering, Louisiana State University, Baton Rouge, LA 70803 USA (e-mail: morooj1@lsu.edu; rsoosa1@lsu.edu; naraghi@lsu.edu). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TVT.2011.2166283 PU is not active. The difficulty is that these requirements must be guaranteed at very low signal-to-noise ratios (SNRs), in the presence of channel impairments and with low computational complexity. Spectrum sensing has received a great deal of attention in recent years, and many algorithms have been proposed. A survey of some recent algorithms appears in [2] and [3]. Noncooperative spectrum-sensing techniques may be classified into two categories, i.e., blind detection and feature detection. Blind detection algorithms assume minimal prior information of the primary signal parameters (e.g., bandwidth and mod- ulation scheme), noise, or the channel. On the other hand, feature detection techniques utilize specific characteristics of the primary signal, noise, and/or channel parameters. Clearly, as more knowledge of the primary signal, noise, or channel is assumed, better performance can be achieved at the expense of additional complexity and less generality. Energy detector (ED) [4], [5] is an example of a sensing technique that is easy to implement. However, selection of a threshold to achieve a given false alarm probability requires precise knowledge of the noise power. In the presence of noise power uncertainty, the performance of ED is severely degraded due to a phenomenon referred to as “SNR wall” [6]. Recently, several covariance and autocorrelation-based al- gorithms have appeared in the literature, which are robust to noise uncertainty and exploit the fact that, while white noise samples are uncorrelated, most communication signals exhibit nonzero correlation. However, these algorithms either assume a specific modulation scheme [7], [8] or rely on oversampling (with respect to the modulation symbol period) of the received signal [9]–[11]. A covariance-based method using the maxi- mum and minimum eigenvalues of the covariance matrix is introduced in [12]. Feature detection techniques include the cyclostationary detectors, which require more information on the primary signal parameters and are more computationally intensive [13]–[17]. Multiple-antenna systems have been widely used in wireless communication to increase channel capacity or to improve transmission reliability. Recently, several authors have con- sidered using multiple-antenna systems for spectrum sensing [5], [11], [18]–[21]. To avoid the problem of noise power uncertainty in EDs, a generalized likelihood ratio test (GLRT) algorithm may be employed [15], [22]. However, the perfor- mance of such algorithms at low SNR values are generally poor. In [19], Taherpour et al. employed GLRT for an ED using a multiple-antenna system and showed that the system is robust to noise power uncertainty. A multiple-antenna system 0018-9545/$26.00 © 2011 IEEE