A Study On The Spectrum Sensing Algorithms Based On Signal Characteristics

Cognitive radio is a promising technique in the wireless communication systems, which can fully utilize the spectrum holes and improve the spectrum efficiency. Secondary users are required to sense the spectrum and opportunistically occupy the idle frequency bands at the time when primary users are not working. Spectrum sensing is a fundamental problem in cognitive radio, which is the process that cognitive users perform signal detecting through a variety of detection methods to obtain the information of spectrum usage at the current time. This paper studies the spectrum sensing techniques of cognitive radio network.The SNR at the secondary user may be very low because of the severe wireless channel condition with multipath fading and shadow fading. Furthermore, the secondary users usually have no knowledge about source signal and the noise. Due to these facts, quick and robust spectrum sensing is much difficult for secondary users. Classic detection methods such as energy detection, likelihood ratio test and matched filtering are not available in cognitive radio systems because they need priori statistical knowledge about source signal and noise. The main task of this paper is to design robust spectrum sensing algorithms with low complexity, which need no knowledge about primary user and noise.In this paper, we propose a spectrum sensing method based on Gerschgorin disk. The Gerschgorin radius contains the information of signal subspace and is robust against non-uniform noise. The proposed method achieves good performance in small sample size, which is important for quick spectrum sensing. Simulation results are included to illustrate the superiority of the proposed method.The asymptotic distributions of test statistic based on Gerschgorin disk are also derived in both signal-absent hypothesis H0 and signal-present hypothesis H1. Simulation results show the error between the derived distributions and real distributions of the test statistic. It is implied that the error between the two distributions becomes small with the increase of SNR and sample size.Orthogonal frequency division multiplexing(OFDM) is a multi-carrier modulation scheme that is widely used in both military and civil digital communication systems. In this paper, spectrum sensing for OFDM signals through multipath channel in cognitive radio system is proposed. The proposed method is derived in the generalized likelihood ratio test(GLRT) framework. Our proposal is compared to several existing OFDM detection methods and is shown to achieve a performance improvement over the existing methods. The computation time of the proposed method is larger than the existing ones, which makes the spectrum sensing more slow. The reduction of computation time of the proposed method is worthy of studying further in the future.