A Statistical Approach to Spectrum Measurement Processing

Prior knowledge of the spectrum behavior can greatly assist in efficiently performing dynamic spectrum access. This knowledge can be obtained by surveying the spectrum activity, which involves spectrum data collection, primary signal detection, and data analysis. One approach to signal detection is to perform energy detection based on the power measurements. Although considered optimal, this approach requires prior knowledge of the noise statistics in order to compute the signal threshold. In this paper, we present several algorithms for estimating the threshold directly from the spectrum measurement data. Both Otsu’s and the recursive one-sided hypothesis testing (ROHT) algorithms, as well as the proposed recursive Otsu’s algorithm, have been applied to spectrum measurements collected from an FM broadcast band and a digital television band. The results show that the proposed algorithm performed well, with nearly 100% detection probability, when applied on the DTV band measurements. However, it was not able to provide the same performance for the FM band measurements.