Minimum error entropy with affine projection algorithm for robust adaptive filtering

Huaiyuan Zhang, Guoliang Li, Yunxian Hou, Hongbin Zhang, Shan Zhong

Abstract: Considering the similarity and correlation between the method of affine projection (AP) and minimum error entropy (MEE), we find that AP and MEE can be combined well, so a new robust adaptive filtering algorithm named APMEE was proposed. By optimizing the non-parametric estimator of quadratic Renyi's entropy of two individual errors in an observation interval with a ℓ2-norm constraint on the weight vector, APMEE greatly improves the convergence speed and steady-state accuracy of adaptive filtering. Numerical simulation experiments of system identification and echo cancellation indicate that APMEE outperforms other algorithms based on affine projection.