(ALMOST) PERIODIC MOVING AVERAGE SYSTEM-IDENTIFICATION USING HIGHER-ORDER CYCLIC-STATISTICS

This correspondence addresses the problem of (almost) periodic moving average (APMA) system identification. Two new normal equations relatin g the coefficients of an APMA system and the time-varying higher order cumulants of the measurements are established, from which two new lin ear algebraic algorithms are presented for system parameter estimation . In addition, a new singular value decomposition (SVD) based algorith m is proposed for determining the system order. Simulation examples ar e given to demonstrate the performance of these new approaches.