On-line statistical processing of radiation detector pulse trains with time-varying count rates

Statistical analysis is of primary importance for the correct interpretation of nuclear measurements, due to the inherent random nature of radioactive decay processes. This paper discusses the application of statistical signal processing techniques to the random pulse trains generated by radiation detectors. The aims of the presented algorithms are: (i) continuous, on-line estimation of the underlying time-varying count rate θ(t)θ(t) and its first-order derivative dθ/dtdθ/dt; (ii) detection of abrupt changes in both of these quantities and estimation of their new value after the change point. Maximum-likelihood techniques, based on the Poisson probability distribution, are employed for the on-line estimation of θθ and dθ/dtdθ/dt. Detection of abrupt changes is achieved on the basis of the generalized likelihood ratio statistical test. The properties of the proposed algorithms are evaluated by extensive simulations and possible applications for on-line radiation monitoring are discussed.