Reduction of errors in blind identification of FIR systems under order overestimation

In this paper, a new algorithm for blind FIR channel identification is proposed. The third- and fourth-order cumulants are used for robustness to Gaussian noise. The proposed algorithm solves an identification equation of channel parameters and estimated cumulants by the gradient descent method. In addition, to solve a singular problem which arises when the channel length is over-estimated, the distance of variables which consists of channel parameters are included in a cost function. Simulation results demonstrate that the performance of the proposed algorithm is superior to that of the previous methods, especially when the length of the FIR channel is over estimated