Variable Step-size Based Adaptive Noise Subspace Estimation for Blind Channel Estimation

The recently proposed HFRANS is a computationally efficient and numerically stable adaptive algorithm for noise subspace estimation. To improve the convergence and steady-state performances of HFRANS, this paper proposes a variable step-size strategy, while maintaining the overall computational complexity at O(NP). The proposed strategy is obtained by cleverly combining a gradient step-size adaptation approach with the optimum step-size, which are developed in this paper. Simulation results are provided to illustrate the performance advantages of the proposed strategy. We also study the application of HFRANS with variable step-size to blind channel estimation in MC-CDMA system