Digital-analog in band separation using extended Combined Measurement Kalman filter

This paper aims at solving the problem of signals separation in the real in-band mixed broadcast of both digital and analog signals. Since the power of OFDM signals is generally much lower than FM signals in mixed signals, it will be destructive to digital signals if digital and analog signals are separated directly, therefore, a unique preprocessing structure is applied to convert digital-analog mixed signals into co-channel FM signals. As to co-channel FM separation, the extended Kalman filter (EKF) and Viterbi algorithm are commonly employed. However, when carrier frequency of FM signals comes to be equal, the co-channel FM signals will be easily crisscross separated by traditional extended Single Measurement Kalman filter (ESMKF) [1] algorithm due to the uncertainty of the battery model even with high SNR. In this paper, an innovative extended Combined Measurement Kalman filter (ECMKF) algorithm, namely predicting and estimating the original signals with more than one observation from the preprocessing structure, is put forward to avoid the crisscross separation effectively. Simulation results show that, compared with the existing ESMKF and Viterbi [2] algorithm, the innovative ECMKF algorithm gains much better performance in co-frequency FM separation than former ones.