Skip to Main Content
Channel impulse response (CIR) prediction is important because it makes possible the provision of up-to-date channel state information which is essential for coherent detection of transmitted message symbols. Different prediction techniques have been proposed for OFDM systems. These range from the minimum mean square error (MMSE) techniques to adaptive techniques. However, it has been confirmed that the adaptive predictors present better performance than its MMSE counterpart. Besides, the computational complexity of the MMSE class of predictors is more costly than the adaptive predictors. In this paper we propose an improved version of an adaptive normalized least mean square (NLMS) predictor named variable step size normalized least mean square (VSSNLMS) predictor. The proposed VSSNLMS predictor is employed for the implementation of decision directed channel estimation (DDCE) for OFDM systems. Simulation results demonstrate that the proposed VSSNLMS predictor outperforms the NLMS predictor at a cost of a negligible high complexity, and its performance is very close to that of the recursive least square (RLS) predictor that exhibits an enormous computational complexity.