FM interference mitigation in a finitely correlated environment using a decorrelating time-varying autoregressive model

A linear prediction error filter (PEF) formed from the coefficients of a time-varying autoregressive (TVAR) model can be used effectively for mitigating frequency modulated (FM) components in a minimally correlated desired signal and white noise. However, when the desired signal is a finitely correlated signal-such as a continuous phase modulated (CPM) signal-the linear PEF formed from the TVAR coefficients results in mitigation of the desired signal along with the interference. We propose a linear PEF formed from the coefficients of a decorrelating TVAR (DTVAR) model to address this issue. A time-varying decision aided compensator (TVDAC) is introduced to compensate for the distortion in the desired signal produced by the proposed PEF. Comparison of a time-varying linear PEF (TVLPEF) formed from DTVAR and TVAR coefficients for interference mitigation in a CPM communications system reveals notable performance gains at most moderate to high signal-to-interference ratios (SIRs).