FIR system identification via second-order statistics
W. Li, Joe C. H. Poon, Wan-Chi Siu · 2002
In many practical applications, it is of interest to identify unknown system characteristics so as to recover input source signals from the observed data. In this paper, we specially consider a model with two receivers, and the outputs are described by FIR equations. Our approach involves the estimation of the FIR coefficients and the construction of the input signals by second-order statistics, i.e. the use of the auto-power and crosspower spectra of the two receiver signals. To accommodate time-varying situations and to serve online purposes, a recursive weighted least squares algorithm is proposed. Comparing with other higher-order statistics approaches, our method is not restricted to independent and identically distributed (i.i.d.) random signals and has less computation burden. Our simulation results also show that the performance of our algorithm is comparable to Giannakis's third-order statistics approach.