On the use of a linear phase lattice structure in Prony´s method for spectral line estimation

It is known that Prony´s method for spectral line estimation requires a filter coefficient symmetric property. Taking advantage of this property, the authors show that the frequencies of the spectral lines can be determined from the reflection coefficients of a cascaded orthogonal FIR (finite-impulse response) lattice algorithm with the data processed in an adaptive mode. An adaptive algorithm based on the LMS criterion has been derived. Simulations indicate that this algorithm can be applied in the estimation of spectral lines in noise. The proposed approach has the advantage of the robustness of lattice modeling techniques, and thereby avoids the numerical problems usually associated with Prony´s method.<>