Title :
Performance bounds for nonlinear filters
Author :
Lu, Shan ; Doerschuk, Peter C.
Author_Institution :
Consat, Clarksburg, MD, USA
Abstract :
An error in a well-known lower bound on the performance of nonlinear fibers based on dynamic statistical models is described. The error occurs in the final step of the proof and deletion of this step results in a valid lower bound which, however, requires evaluation of a conditional expectation in function space. For the limited, but interesting, class of linear Gaussian process models, computational methods are described.
Keywords :
Gaussian processes; circuit analysis computing; error analysis; nonlinear filters; performance evaluation; statistical analysis; computational methods; dynamic statistical models; error; linear Gaussian process models; nonlinear filters; performance bounds; Computational modeling; Concrete; Error correction; Gaussian processes; Integral equations; Mean square error methods; Monte Carlo methods; Nonlinear filters; Random variables; Rate distortion theory; Rate-distortion;
Journal_Title :
Aerospace and Electronic Systems, IEEE Transactions on