Your browser does not support JavaScript!
http://iet.metastore.ingenta.com
1887

access icon free Multi-stage least mean square algorithm based on polynomial fitting in time-varying systems

A multi-stage least mean square (MLMS) algorithm based on polynomial fitting in time-varying systems is proposed. As the name implies, the MLMS algorithm is divided into different stages in sequence. In non-stationary environments, the optimal weight coefficients of the adaptive filter are variable with time and they are assumed to be approximated by a polynomial. According to the conditions above, the convergence of the MLMS algorithm is verified through theoretical analysis and formula derivation. Similar to the least mean square (LMS) algorithm, the step size μ must be chosen in the specific range to guarantee the feasibility and effectiveness of the MLMS algorithm. Compared with the LMS, LMF and NLMS algorithms, the MLMS algorithm provides the best convergence performance in the simulation experiments.

References

    1. 1)
    2. 2)
      • 4. Nunoo, S., Chude-Okonkwo, U.A.K., Ngah, R.: ‘Performance of LMS, NLMS and LMF algorithms in tracking time-varying UWB channels’. 2013 IEEE Int. Conf. Signal and Image Processing Applications (ICSIPA), Melaka, Malaysia, October 2013, no. 1, pp. 312316.
    3. 3)
      • 2. Eweda, E.: ‘Tracking analysis of the normalized LMS algorithm without the independence and small step size assumptions’. 2009 IEEE Int. Symp. Signal Processing and Information Technology (ISSPIT), Ajma, United Arab Emirates, December 2009, pp. 129134.
    4. 4)
      • 3. Haykin, S.: ‘Adaptive filter theory’ (Prentice-Hall, Upper Saddle River, NJ, USA, 2002, 4th edn.).
    5. 5)
      • 5. Diniz, P.S.R.: ‘Adaptive filtering: algorithms and practical implemen- tations’ (Kluwer, Boston, MA, 2002, 2nd edn.).
http://iet.metastore.ingenta.com/content/journals/10.1049/el.2017.4330
Loading

Related content

content/journals/10.1049/el.2017.4330
pub_keyword,iet_inspecKeyword,pub_concept
6
6
Loading
This is a required field
Please enter a valid email address