Reweighted lp-norm constraint least exponentiated square algorithm for sparse system identification

A reweighted l(p)-norm constraint least exponentiated square (RLP-LE2) algorithm is proposed in this paper. The RLP-LE2 algorithm is derived by minimizing a differentiable cost function composed of the exponential error, l(p)-norm of the weight vector, and reweighted zero attractor. We also replace the l(p)-norm of the weighted factor with an approximation to reduce computation complexity. As we know, the algorithms with the zero attractor are designed to identify spare system. So simulations conducted in the system identification are provided to demonstrate the effectiveness of the proposed algorithm. In this paper the proposed algorithm achieves a low steady-state misalignment and robustness under the impulsive environments comparing the performances of this and several exiting algorithms.