Nonlinear estimation using correlation information

The approximate Karhunen-Loegraveve (KL) expansions are used as a tool for systems identification providing an approximate state-model for the signal of interest. The identified model can then be used to derive sequential algorithms for the nonlinear estimation problem by using standard estimation techniques. The proposed solution is specially indicated in those situations where a state equation is not at hand and the only information available is the correlation matrix of the signal