Recursive Parameter Estimation for Regression Channel Model in Pilot-Aided OFDM Systems

OFDM model-based channel estimation techniques conventionally apply the least squares method to estimate parameters in a regression model through uniformly distributed pilots in a local region. However, the model estimation must use as many pilots as possible to reduce the effect of noises with the penalty of increasing the storage size for received OFDM symbols. We observed that the model parameters between neighboring local regions are correlated. Hence, some recursive methods are proposed to adaptively estimate model parameters such that the required number of pilots can be reduced, and thus, the required storage size is reduced for interpolating the symbols used in the regression model. Theoretical analysis and simulations show that a better performance is obtained as well by using the proposed parameter estimation methods.