Blind Spectrum Sensing Based on Petrosian's Algorithm in Frequency Domain

Abstract

Spectrum sensing is the key and premise of cognitive radio. The current spectrum sensing methods have some failures and limitations, such as sensitivity to noise uncertainty and failure in some special modulation types. In order to cope with these problems and improve the detection performance, we propose a blind spectrum sensing method based on Petrosian's algorithm in frequency domain. It calculates the fractal dimension by Petrosian's method in frequency domain, and compares the fractal dimension with the threshold to determine whether the primary users exist or not. Monte Carlo simulations show that the proposed method can blind sense spectrum effectively. It outperforms energy detection, box dimension method and Higuchi dimension method with about 1 dB, 25 dB and 25 dB improvements respectively, and consumes much shorter detection times than them. Furthermore, it can overcome some failures and limitations of some other spectrum sensing methods, such as sensitivity to noise uncertainty and failure in some special modulation types, and can be applied to fast blind spectrum sensing in situations of noise uncertain.