A Variable Parameter LMS Algorithm Based on Generalized Maximum Correntropy Criterion for Graph Signal Processing

• Published in: IEEE transactions on signal and information processing over networks Vol. 9; pp. 140 - 151
• Main Authors: Zhao, Haiquan, Xiang, Wang, Lv, Shaohui
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adaptive filters; Algorithms; Criteria; Fourier transforms; generalized maximum correntropy criterion; graph fourier transform; Graph signal processing; impulse noise; Information processing; Mean square error methods; parameter optimization; Parameters; Robustness; Sampling; Signal processing; Signal processing algorithms; Steady state
• ISSN: 2373-776X; 2373-7778
• DOI: 10.1109/TSIPN.2023.3248948

The least mean square (LMS) algorithm of the graph signal processing (GSP) based on the mean square error criterion has a poor reconstruction effect when the graph sampling signal is disturbed by impulse noise. To solve this problem, the generalized maximum correntropy criterion (GMCC) is introduced, which is robust to impulse noise in adaptive filtering. Therefore, this paper proposes the GSP LMS algorithm based on the GMCC (GSP LMSGMCC) by using the graph Fourier transform, which has a good effect when the graph sampling signal is disturbed by impulse noise. In addition, the GSP LMSGMCC algorithm based on the fixed parameter including step size and kernel width must make a compromise between convergence speed and steady-state error. To prevent this, the fixed parameters of the proposed GSP LMSGMCC algorithm are optimized, respectively. To facilitate understanding and analysis, the steady-state performance of the proposed GSP LMSGMCC algorithm is studied. Finally, the computer simulations are carried out to verify the superiority of the proposed algorithm when the signals on the graph are static graph signals and streaming graph signals respectively.