Open AccessReweighted l p constraint LMS‐based adaptive sparse channel estimation for cooperative communication system
Author(s) -
Zhang Aihua,
Liu Pengcheng,
Ning Bing,
Zhou Qiyu
Publication year - 2020
Publication title -
iet communications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.355
H-Index - 62
eISSN - 1751-8636
pISSN - 1751-8628
DOI - 10.1049/iet-com.2018.6186
Subject(s) - norm (philosophy) , channel (broadcasting) , computer science , least mean squares filter , system identification , algorithm , mathematical optimization , constraint (computer aided design) , compressed sensing , convergence (economics) , mathematics , control theory (sociology) , adaptive filter , telecommunications , artificial intelligence , data modeling , geometry , control (management) , database , political science , law , economics , economic growth
The issue of sparsity adaptive channel reconstruction in time‐varying cooperative communication networks through the amplify‐and‐forward transmission scheme is studied. A new sparsity adaptive system identification method is proposed, namely reweighted l p norm ( 0 < p < 1 ) penalised least mean square (LMS) algorithm. The main idea of the algorithm is to add a l p norm penalty of sparsity into the cost function of the LMS algorithm. By doing so, the weight factor becomes a balance parameter of the associated l p norm adaptive sparse system identification. Subsequently, the steady state of the coefficient misalignment vector is derived theoretically, with a performance upper bounds provided which serve as a sufficient condition for the LMS channel estimation of the precise reweighted l p norm. With the upper bounds, the authors prove that the l p ( 0 < p < 1 ) norm sparsity inducing cost function is superior to the reweighted l 1 norm. An optimal selection of p for the l p norm problem is studied to recover various d sparse channel vectors. Several experiments verify that the simulation results agree well with the theoretical analysis, and thus demonstrate that the proposed algorithm has a better convergence speed and better steady‐state behaviour than other LMS algorithms.