A Variable Step-Size Affine Projection Algorithm Based on Data Reuse

The affine projection algorithm (APA) provides good performance features, especially for correlated input signals. For further improving its performance, the step-size parameter of the algorithm can be controlled to better control the opposite requirements of fast convergence/tracking and low misadjustment (i.e., good accuracy). In this work, we exploit the data-reuse method in conjunction with the APA. Using this method, a sequence of step-size parameters can be scheduled a priori and used within the APA, so that the resulting algorithm acts like a variable step-size version. Experiments performed in the framework of acoustic echo cancellation support the practical features of the proposed solution.