Constrained Least Mean M-Estimation Adaptive Filtering Algorithm

• Published in: IEEE transactions on circuits and systems. II, Express briefs Vol. 68; no. 4; pp. 1507 - 1511
• Main Authors: Wang, Zhuonan, Zhao, Haiquan, Zeng, Xiangping
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.04.2021; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adaptive algorithms; Adaptive filters; Adaptive systems; Algorithms; Circuit stability; Complexity; Computational complexity; Constrained adaptive filtering; Cost function; Filtering algorithms; M-estimate; Mathematical model; non-Gaussian noises; Random noise; Stability analysis; Steady state; system identification
• ISSN: 1549-7747; 1558-3791
• DOI: 10.1109/TCSII.2020.3022081

In many applications, the constrained adaptive filtering algorithm has been widely studied. The classical constrained LMS algorithm is widely used because of its low computational complexity. However, the performance of constrained LMS algorithm will degrade under correlated input or non-Gaussian noise. In order to overcome this defect, this brief proposes a constrained least mean M-estimation (CLMM) algorithm, which uses the M-estimation cost function for the constrained adaptive filter. Compared with the previous algorithms for non-Gaussian noise, such as constrained maximum correntropy criterion (CMCC) algorithm and constrained minimum error entropy (CMEE) algorithm, the proposed CLMM algorithm has lower computational complexity and better steady-state performance. In addition, the step-size range is determined by analyzing the mean square stability, which ensures the stability of the proposed CLMM algorithm. Simulation results illustrate that the proposed CLMM algorithm has better steady-state performance than previous algorithms in non-Gaussian noises with multi-peak distribution.