Subspace Tracking-Based Decision Directed CIR Estimator and Adaptive CIR Prediction

Abstract

One of the essential prerequisites for coherent detection in Orthogonal Frequency Division Multiplexing (OFDM)-based communication systems is the availability of an accurate estimate of the channel state information (CSI). In the absence of symbol errors, the decision directed channel estimation (DDCE) scheme has been proved to provide better and more accurate CSI than the channel estimation scheme based purely on pilot symbols. In this paper a subspace algorithm, the fast data projection method (FDPM), is employed for estimation of the channel impulse response (CIR) of the OFDM channel in a decision directed mode. A variable step size normalized least mean square (VSSNLMS)-based predictor is derived for the implementation of the CIR prediction stage of the estimation scheme. The whole channel estimation scheme based on the two algorithms is simulated in the context of fractionally spaced (FS)-CIR based channel model, also known as non-sample spaced CIR model. The performance of the FDPM-based decision directed channel estimator is compared through computer simulation with another subspace algorithm, deflated projection approximation subspace tracking (PASTd)-based channel estimator, while employing normalized least mean square (NLMS) and the proposed VSSNLMS predictors, respectively. It was discovered from the simulation results that the proposed scheme performed better than the PASTd-based channel estimator. The channel estimation scheme employing VSSNLMS-predictor also shows enhanced performance in comparison with the scheme using NLMS-based predictor.