Bayesian Wideband Spectrum Segmentation for Cognitive Radios

A novel approach for wideband spectrum segmentation is proposed. The proposed method, called segmented periodogram, is based on the posterior expectation of the piece-wise flat realizations of the underlying signal spectrum, which is obtained through the simulated samples drawn from the joint posterior using the reversible jump Markov chain Monte Carlo (RJMCMC) technique. The segmented periodogram is a coarse description of the spectrum, which can be further utilized to obtain desired spectral parameters. The proposed framework is suitable for cognitive radios (CRs) as it allows the use of knowledge built from past experiences in the form of marginal densities as prior initializations in future calculations.