Connections between classical and statistical methods of FIR digital filter design

Consider the weighted squared error between an FIR filter frequency response and a desired frequency response. The FIR filter coefficients that minimize this error are shown to satisfy a linear Toeplitz set of equations. Thus, the MMSE designs of [1] may arise also as solutions to a classical filter minimization problem. The solutions are linear phase, and identical with those of [1] when the desired frequency response and the weighting function have a special form.