Applying Csiszár's I-divergence to blind sparse channel estimation

Compressed sensing (CS) has renewed interest in sparse channel estimation. Herein, a semi-blind, iterative, sparse channel estimation method is proposed. The new method is based on minimizing Csiszár’s I-divergence using Schulz & Snyder’s iterative deautocorrelation algorithm. First, it is shown that the desired methods can be adapted to the problem of interest. The proposed semi-blind method accurately estimates the significant tap locations of a sparse channel, and their corresponding magnitudes. A method for determining the channel coefficients up to a phase ambiguity is presented. The simulation results show that although limited pilots are used, the proposed semi-blind iterative algorithm achieves performance comparable to that of training-based compressed sensing methods.