Exploiting Interference Diversity for Event-Based Spectrum Sensing

Abstract

Abstract-Spectrum sensing is a core problem in cognitive radio. Detecting the presence/absence of very weak primary users with a single antenna can be very difficult. Earlier it have been shown that uncertainties in the environment result in SNR walls that detectors cannot beat in a robust manner. Multiple antenna approaches show the potential of getting gains, but we show here that for a single user, multiple antenna detection still must suffer from an SNR Wall. The reason is that the real world uncertainty in noise is dominated by the potential presence of an unknown number of low-powered interference sources in the external environment. The traditional approach to collaborative spectrum sensing attempts to use the shadowing/multipath diversity across different users to boost the reliability of detection. We show here that there is another kind of diversity that is also available: interference diversity. This diversity captures the fact that these low-powered interference sources are local to individual users whereas the primary user has a global footprint. To exploit this diversity, we must shift our perspective from existence-based detection (whether the primary is present or not) to event-based detection (whether the primary has turned off or on). We study this and explore the limits to this approach.

Cooperative robust sequential detection algorithms for Spectrum Sensing in Cognitive Radio

“…Because cognitive radios have to detect primary signals at very low SNR, these modeling uncertainties can cause an SNR wall, a level below which spectrum sensing fails to be robust to modeling uncertainties. More recently it is shown in [17] that if the primary user goes ON and OFF at different times one may be able to detect this change in a reliable and robust manner and then SNR wall can be breached. Of course in this case one is looking for spectral holes in time.…”

Sequential detection based cooperative spectrum sensing algorithms in cognitive radio

“…Change time T is fixed at 200. The results are shown in Finally we compare the MGLR-CUSUM algorithm to the algorithm in [15] which is actually a non-parametric cooperative algorithm. There are 5 nodes and the SNR in each node is -3dB.…”

“…In this algorithm instead of CUSUM, GLR ( [11]) is used at the local nodes, which also has some optimality properties. We will show that our algorithm performs better than the algorithms in [10] and [15].…”

“…There are 5 nodes and the SNR in each node is -3dB. The SNR is kept same at each node to facilitate easier comparison with the algorithm in [15] …”

“…Because cognitive radios have to detect primary signals at very low SNR, these modeling uncertainties can cause an SNR wall, a level below which spectrum sensing fails to be robust to modeling uncertainties. More recently it is shown in [15] that if the primary user goes ON and OFF at different times one may be able to detect this change in a reliable and robust manner and then SNR wall can be breached. Of course in this case one is looking for spectral holes in time.…”