A Frequency Domain State-Space Approach to LS Estimation and Its Application in Turbo Equalization

Abstract

A frequency domain state-space approach to the least squares (LS) estimation for finite impulse response (FIR) system identification is proposed by exploiting the Toeplitz structure of the data matrix, leading to two low-complexity FFT (fast Fourier transform)-based unbiased recursive estimators. The proposed recursive estimators are applied to the smoothing estimation of timevarying frequency selective channels in turbo equalization systems, and channel estimation approaches with complexity (log ) per symbol are developed, where is the number of channel taps.