Error Bounds and Asymptotic Performance under Mismatch of Multisensor DetectionSystems

摘要

A binary detection problem, when the data are obtained distant sensors, andmodelled as stationary Gaussian processes, with different spectra is considered. It is also assumed that inaccurate versions of the true statistical models are utilized, and upper bounds to the probability of error, based on the Chernoff bounding approach are developed. Those bounds also converge to the asymptotic probability of error as the number of data points increased. Conditions for sustaining, in spite of mismatch, exponential convergence of the error probability to zero are also determined. (rh)