Steady-State Mean-Square Error Performance Analysis of the Tensor LMS Algorithm

Ningning Zhang, Jingen Ni, Jie Chen, Zhe Li

科研成果: 期刊稿件文章同行评审

20 引用 (Scopus)

摘要

In some system identification problems, the weight vector to be estimated can be represented by a tensor product of two low dimensional component vectors. This feature is useful for the design of adaptive filters with fast convergence rate and/or low computational complexity. The tensor LMS algorithm has been proposed under this scenario and its recent extensions find applications in several contexts. This brief analyzes the steady-state mean-square error (MSE) performance of the tensor LMS algorithm, which can provide performance prediction and design guideline for engineers. Since the updates of the weight-vectors of the component adaptive filters (CAFs) are coupled, it is challenging to perform steady-state MSE analysis. To address this problem, this brief first establishes the variance relation in steady state for the CAFs and then decouples their excess MSEs by solving a system of linear equations. Finally, the steady-state MSE of the reproduced adaptive filter is composed of the excess MSEs of the CAFs and the variance of the system noise. Simulations are performed to verify the accuracy of the theoretical findings.

源语言英语
文章编号9177175
页(从-至)1043-1047
页数5
期刊IEEE Transactions on Circuits and Systems II: Express Briefs
68
3
DOI
出版状态已出版 - 3月 2021

指纹

探究 'Steady-State Mean-Square Error Performance Analysis of the Tensor LMS Algorithm' 的科研主题。它们共同构成独一无二的指纹。

引用此