Performance Evaluation of Energy Detectors in Cognitive Radio

Abstract

Wireless networks are characterized by a fixed spectrum assignment policy. With the advancement of wireless communications, the problem of bandwidth scarcity has become more prominent. As per Federal Communications Commission (FCC), large portion of the spectrum lies vacant most of the time and that portion is the licensed spectrum band; which is utilized by licensed users only. The temporal and geographical variation in the usage of the assigned spectrum ranges from 15% to 85\%. So, to solve this problem of spectrum under-utilization, FCC allowed secondary users to utilize the licensed band when it is not in use and named it as Cognitive Radio.

To sense the existence of licensed users or in other words, to utilize the unused spectrum, spectrum sensing techniques are used. Energy detection, Matched filter detection and Cyclo-stationary feature detection are the three conventional methods used for spectrum sensing. Each technique has its own advantages and drawbacks. Matched filter spectrum sensing technique requires a priori information about each primary user and a dedicated cognitive radio receiver is required for every primary user. Cyclostationary feature Detection is computationally complex and requires significantly long observation time to extract the features of primary user signal. Energy detection is the most simplest to implement, but the performance of energy detector is susceptible to uncertainty in noise power. This report discusses the conventional energy detection method in case of AWGN channel and Rayleigh fading channel. The performance is improved by introducing diversity in fading channels. The generalized energy detection method is discussed where squaring operation of conventional energy detection is replaced by any positive power constant, which is known as generalized energy detector. Also, effect of noise uncertainty is studied in this generalized energy detector. The performance of energy detector degrades significantly under low SNR circumstances and detection becomes impossible below certain critical values called “SNR Walls”. To improve the detection probability under such case, Stochastic Resonance (SR) based energy detection approach is used. It significantly reduces the SNR wall. A novel combination of generalized energy detector and SR phenomenon is presented here and improved results have been obtained. Mathematical Analysis has been illustrated for all these cases. Simulation and analytical results have also been included in this report.