A Cooperative Spectrum Detection Technique in Non-Gaussian ...


A Cooperative Spectrum Detection Technique in Non-Gaussian ...



Simulations are carried out with σ


= 1, σ


= 81 , and ε = 0. 3 corresponding to

2 '

σ = 25 and I


≈ 0.56 . When given false alarm probability P

f a

is 0.01 and 0.03, the threshold γ is

6.635 and 4.709, respectively. By using the threshold γ '

and non-centrality parameter λ , the

detection probability can be derived, as shown in Fig.3. We can see that the detection probability is

a monotonically increasing function of λ (or SNR). For the mixed-Gaussian Rao


detection λ = SNR(

σ I


) , while for the Gaussian Rao detection, λ = SNR . With the same signal and

non-Gaussian interference backgrounds, the mixed-Gaussian Rao detection has gained an

additional SNR improvement about 11dB, which is in agreement with the theoretical value:

Δ= I = dB



10lg( σ


) 11.4

4.2 Cooperative Detection and Non-Cooperative Detection

Suppose three users CU1, CU2 and CU4 in Fig.1 are chosen to cooperate, and their SNRs are set to

-3dB, -5dB, -10Db, respectively. The resulting ROC curves are shown in Fig.4. Obviously, under

the same probability of false alarm conditions, the probability of detection is increased as the SNR


Figure 4. Non-cooperative detection and traditional OR cooperation

Figure 5. Proposed cooperation based on Rao detection

Fig.4 also indicates that when multiple users cooperate and the SNR of CU2 is very low, the

probability of cooperative detection is less than the probability of detection by CU4 alone if using


More magazines by this user
Similar magazines