Determining the geographical area that needs to be excluded due to incumbent activity is critical to realize high spectral utilization in spectrum sharing networks. This can be achieved by estimating the incumbent location and transmit power. However, keeping the hardware complexity of sensing nodes to a minimum and scalability are critical for spectrum sharing applications with commercial intent. We present a discrete-space l1-norm minimization solution based on geolocation-aware energy detection measurements. In practice, the accuracy of geolocation tagging is limited. We capture the impact as a basis mismatch and derive the necessary condition that needs to be satisfied for successful detection of multiple incumbents' location and transmit power. We find the upper bound for the probability of eliminating the impact of limited geolocation tagging accuracy in a lognormal shadow fading environment, which is applicable to all generic I1-norm minimization techniques. We propose an algorithm based on orthogonal matching pursuit that decreases the residual in each iteration by allowing a selected set of basis vectors to rotate in a controlled manner. Numerical evaluation of the proposed algorithm in a Licensed Shared Access (LSA) network shows a significant improvement in the probability of missed detection and false alarm.
- cognitive radio
- Licensed Shared Access (LSA)
- orthogonal matching pursuit (OMP)
- sparse matrices
- spatial correlation