【正文】 by the sum criteria, where M is the number of random variables, and meets the χ2distribution.Pd=1Q(λi=1Lμ(τs(i))2i=1Lμ(τs(i)))The simulated signal abovementioned is detected circularly, then the coordinate axes are normalized and the results are as follows.Figure 4. Twodimensional autocorrelation function of signalFigure 5. Relationship between delay and autocorrelation functionThe two pictures above are twodimensional picture of autocorrelation function of the signal RXατ, varied from cycle frequency α and delayτ and α = 0 crosssection picture. We can see that RXατ exists nonzero values in certain loop frequency and time delay, especially in the twice carrier existing the peak. Fig. 5 presents the relationship between delay and autocorrelating function, which help analyzing and selecting delay parameters τs to obtain detecting probability. Cyclostationary detection is in cycle frequency domain and the phase and frequency information relevant with time parameters of OFDM signal is retained in signal processing, thus overlapping features of power density in the frequency domain is no longer overlap in the loop. Both signal and noise has spectral ponents when α = 0, while the spectrum ponent of noise is zero when α ≠ 0 , and the zero cycle frequency corresponds to the smooth part of signal, only nonzero cycle frequency portray the cyclostationarity of signal.Figure 6. Cycle spectrum of simulated signalFigure 7. α= 0, aspect picture of OFDM spectral correlation functionFrom the figures above, we can see that the spectral characteristics of OFDM signal can be reacted truly by the method of cyclic spectrum related. OFDM signal shows cyclostationary characteristics, while the stationary noise does not show the relevance characteristics of cyclic frequency domain, and interference signals usually shows cyclostationary characteristics different from the signal of main users. Therefore, the greatest feature of cyclostationary detection has the ability to distinguish the spectrum from the primary user signal, noise and interference signal. We only need to verdict whether the spectrum lines appears when α ≠ 0 as the verdict condition, which is an important way of cyclostationary detection. Detected OFDM signal by cyclostationary detection can get rid of the influence of background noise, and distinguish the noise energy with the main user signal energy. Thus, detection performance is still good even in low SNR. The detection performance of cyclostationary detection is superior to energy detection by paring the detection probability of them.Figure 8. Comparison between cyclostationary detection and energy detectionIV. CONCLUSIONClassic cognitive radio frequency spectrum detection method such as energy detection and cyclostationary detection in OFDM system is still feasible to detect the spectrum. The algorithm of energy detection is simple and is easy to implement。 if there are authorized users in the channel, perceived signal is posite signal of noise signal and certain signal, whose mean is nonzero, and thus subjects to noncentral χ2 distribution, the formula is as follows.S~χ2L2 , H0χ2L2(λ), H1Detection probability Pdand false alarm probability Pd of two assumptions are as follows.Pd=PSλH0=Γ(L,λ/2)LPf=PSλH1=QL(2SNR,λ)where SNR is the signal to noise ratio. A string of OFDM si