Recursive Channel Estimation for Turbo Equalization Based on a State-Space System Model

In this paper, we first develop a state-space system model for channel estimation, then propose a recursive channel estimation approach using the Gaussian message passing (GMP) technique, which is equivalent to the LS (least squares) approach. By introducing two different approximations to a GMP updating rule, the computational complexity of this recursive approach is significantly reduced, leading to two low-complexity unbiased FFT (fast Fourier transform)-based channel estimators. The proposed estimators are very suitable for turbo equalization, where the estimates of the data symbols available from the decoder are exploited as a virtual training sequence for channel estimation (In contrast, a direct use of the LS estimator incurs high complexity since the matrix inversion involved needs to be calculated online). We apply the proposed channel estimators in turbo equalization of quasi-static and time-varying frequency selective channels, and simulation results demonstrate their effectiveness.