Two-stage common error hierarchical normalized least-mean-square (NLMS) algorithm is presented in the context of network echo cancellers and sparse systems. The suggested adaptive filter structure is generic, uses a common error feedback for both stages, and is applicable with any type of error minimization technique. The simulation results show that the two-stage method exploits the sparseness of the system better than the proportionate NLMS (PNLMS) while keeping the initial convergence rate intact and improving the steady state convergence time significantly.
展开▼