Optimal Step-Size Control of the Partitioned Block Frequency-Domain Adaptive Filter

The partitioned block frequency-domain adaptive filter (PBFDAF) is appealing in echo cancellation due to its computational efficiency and good convergence property. However, the choice of the step size should comprise between the convergence rate and the final misalignment. Several variable step-size schemes were proposed, but this problem was not resolved entirely. In this brief, a robust optimal control approach to the PBFDAF algorithm is presented. We carry out a statistical convergence analysis for the PBFDAF, which is then used to derive the optimal step size and estimate the system distance. Simulation results demonstrate that the proposed method achieves an improved performance than the other variable step-size PBFDAF algorithms in the context of echo cancellation.