Combined Spectrum Sensing Technique for Cognitive Radios

Abstract

As more and more wireless systems are being developed in order to operate in crowded spectrum bands, radio spectrum is being used inefficiently or is underutilized. Most of the spectrum has been allocated to specific users, while other spectrum bands that haven’t been assigned are overcrowded because of overuse. However, most of the allocated spectrum is idle sometimes; it has been observed that a large part of the allocated spectrum is underutilized. The solution to this problem is Cognitive Radio. Cognitive radio senses the radio spectrum and collect information from the past behavior of the primary user. It detects the spectrum hole of the same bandwidth as required by the SU and allocates the spectrum dynamically without interfering with primary user. There are many techniques which can be used for spectrum sensing such as Energy detection (ED), Matched Filter Detection (MFD), and Cyclostationary Feature Detection (CFD). Out of these techniques, Energy detection is the best choice for spectrum sensing if no prior knowledge about the signal is required. This technique is easy to implement but is not suitable for low SNR (signal to noise ratio) environment and conditions where noise is uncertain in channel. Matched Filter Detection on other hand can detect signals with low SNR, but it requires prior knowledge about primary user which limits its application. Cyclostationary feature detection is the most suitable choice as compared to ED and MFD. Cyclostationary processes are random process for which the second order statistics such as mean and autocorrelation change periodically with time. Most of the manmade signals are random in nature i.e. they exhibit underlying periodicities in their signal structures and hence can be called as cyclostationary processes. But many of the technologies that are employed for spectrum sensing are based on stationary statistics which uses probabilistic models to analyze noise contaminated communication signals and hence are not appropriate for received manmade signals. In this work, two main techniques namely Energy Detection and Cyclostationary Feature Detection are investigated. Results show that in case of Energy Detection for 10% probability of false alarm, the probability of miss is 21%. Then comparison of Energy detection and CFD is carried out which shows that CFD reduces probability of miss by 88.57% as compared to ED. Then strategy is proposed based on combination of both the viii techniques for spectrum sensing which outperforms both the existing techniques and achieves reduction in probability of miss by 98.33% as compared to ED.