System Identification in GSM/EDGE Receivers Using a Multi-Model Approach

Abstract—Model order selection is an important element in system identification. It is well known that common model order selection methods such as Akaike’s information criterion (AIC) and Bayesian’s information criterion (BIC) neglect relevant information that is available in models of order different from the one chosen. In this paper the model order selection problem for receivers similar to those found in GSM and EDGE systems is reviewed briefly and is solved with a multi-model approach based on simultaneous consideration of several models. Two methods are evaluated; a multi-model noise suppression filter and multi-model soft combining. The algorithms are implemented and evaluated by means of simulations. The performance of each method is analyzed for GSM and EDGE receivers in a link level simulator. Simulation results show a significant improvement in performance at the cost of increased computational complexity for the multi-model approach