A GENERAL RECURSIVE DISCRETE-TIME FILTER

A reference probability is explicitly constructed under which the sign al and observation processes are independent. A simple. explicit recur sive form is then obtained for the conditional density of the signal g iven the observations. Both nonlinear and linear filters are considere d, as well as two different information patterns, AMS 1991 CLASSIFICAT ION: PRIMARY 93 E11 SECONDARY 60 G35