Signal Processing

Abstract

From the last lecture, Burg defined a sequential method to determine “optimal” forward and backward predictors from a given random sequence. These equations take the form

$$ {f_N}(z){\text{ = }}{f_{N - 1}}(z) + {\user1{}_N}z{b_{N - 1}}(z) $$

$$ {b_N}(z) = z{b_{N - 1}}(z) + {\bar \hat c_N}{f_N}(z) $$