A Hierarchial MDL Based Approach for the Selection of Significant Channel Parameters in OFDM Systems

The paper considers a hierarchial minimum description length (MDL) based model selection method for the identification (classification) of significant channel parameters in OFDM systems. Unlike many OFDM channel estimation methods, where the channel parameters are estimated based on a known or an estimated length, in this paper, not only the channel length is assumed to be unknown but also only significant channel parameters will be identified using the proposed method. Firstly, a model is proposed for the channel response. By using a training based scenario, the model parameters are estimated using least squares (LS) or IFFT. Based on those estimates, an MDL model selection method is used to estimate the channel length and classify each significant parameter of the channel response. Simulation results show that the method is capable of identifying (classifying) significant channel parameters with high probability. In addition a receiver based on those parameters have a better performance than the one which based on the full model