A fast Bayesian model for a latent radio signal

Abstract

Abstract?This paper considers the use of a recently developed Bayesian statistical approximation technique that leads to very fast determination of highly accurate estimates for latent radio signal power. Following suitable data analysis, a first order non-stationary auto-regressive process is considered for latent radio signal and the fast approximation technique is then used to provide accurate estimates of the hidden model parameters. These estimates are based on having received several noisy, but spatially correlated, observations of the true latent signal. The implication of this technique for real time decision analysis and the problem of finding, and making use of, so-called radio spectrum holes is also discussed.