Improving accuracy for TVWS geolocation databases: Results from measurement-driven estimation approaches

Recent advances towards opening underutilized spectrum resources for secondary use rely on geolocation databases to determine protection requirements of primary users. This paper studies the achievable accuracy of signal strength estimation for a primary TV network with log-normal propagation models, linear interpolation through Delaunay triangulation, and ordinary kriging as means of including measurement data into a database's prediction process. Direct estimation methods are compared against hybrid mixture models of statistical propagation modelling with interpolation of the error surface. The results presented herein use measurements from a large-scale measurement campaign in the UK. They show that relatively small data sets suffice to achieve mean absolute prediction errors of 3.1-5.0dB compared to 5.9-6.9dB achieved by purely terrain-based estimation approaches.