Revisiting TV Coverage Estimation with Measurement-based Statistical Interpolation

Per FCC rules, secondary users in TV White Spaces must operate only within spectrum subject to a no-harmful-interference condition to existing primary receivers. In effect, this translates into a protection region around every TV transmitter, wherein secondary nodes must not transmit (on the same channel). This is implemented by requiring secondary users to consult a database prior to any channel access so as to obtain the local prohibited channels/spatial zones (or equivalently the free channels or White Spaces). Construction of such protection regions for a transmitter within a database has been done, mainly based on empirical propagation models that estimate the received signal strength at a location. Clearly, such model-based prediction is always of limited accuracy and should be supplemented by measurement based approaches that help validate and improve the database predictions. In this work, we present results based on applying spatial interpolation techniques (Kriging) to measurement data obtained in Seattle, WA. Our results have shown that empirical DBA models tend to over-estimate received signal strengths by not explicitly accounting for local obstructions, while measurement-based Kriging achieves consistently good performance. Furthermore, boundary estimation via Kriging achieves a type I error rate 46.1 % lower than comparable DBA approach while keeping type II error rate under a low limit (5%) for a given service threshold (i.e., -84 dBm/6 MHz); this is also an improvement over a method using k-Nearest Neighbor for such estimation.