A modified variable degree variable step size LMS algorithm

In this paper, the data reusing technique which utilizes the LMS algorithm is further investigated. It is shown that a near-optimal performance can be achieved if the number of data reusing runs, M, is made variable. Two modifications are proposed in this paper. The first one is based on updating ea...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Circuits and Systems; Proceedings: Midwest Symposium on Circuits and Systems s. 506 - 509
Hlavní autoři: Haddad, M.I., Khasawneh, M.A.
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 1998
Témata:
Computer simulation
Convergence
Equations
Error correction
Least squares approximation
Robustness
Steady-state
ISBN:9780818689147, 0818689145
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In this paper, the data reusing technique which utilizes the LMS algorithm is further investigated. It is shown that a near-optimal performance can be achieved if the number of data reusing runs, M, is made variable. Two modifications are proposed in this paper. The first one is based on updating each individual component of the weight vector with a different step size. This makes the modified algorithm more capable of tracking weight variations than the original algorithm. The second modification makes the algorithm behave initially as normalized LMS with unity step and at the steady state the algorithm behaves like the LMS with a constant step size. Computer simulations show that the first modification gives the algorithm the capacity to perform better for correlated signals than the original algorithm. Also the second modification is found to give the algorithm the ability to perform very close to optimal for uncorrelated signals.
ISBN:9780818689147
0818689145
DOI:10.1109/MWSCAS.1998.759541