RLS adaptive filter with inequality constraints

Vitor Nascimento; Yuriy Zakharov

doi:10.1109/LSP.2016.2551468

RLS adaptive filter with inequality constraints

Vitor Nascimento, Yuriy Zakharov

Electronic Engineering

Research output: Contribution to journal › Article › peer-review

Abstract

In practical implementations of estimation algorithms,
designers usually have information about the range in
which the unknown variables must lie, either due to physical
constraints (such as power always being nonnegative) or due to
hardware constraints (such as in implementations using fixedpoint
arithmetic). In this paper we propose a fast (that is, whose
complexity grows linearly with the filter length) version of the
dichotomous coordinate descent recursive least-squares adaptive
filter which can incorporate constraints on the variables. The
constraints can be in the form of lower and upper bounds on each
entry of the filter, or norm bounds. We compare the proposed
algorithm with the recently proposed normalized non-negative
least mean squares (LMS) and projected-gradient normalized
LMS filters, which also include inequality constraints in the
variables.

Original language	English
Pages (from-to)	752-756
Number of pages	5
Journal	IEEE Signal Processing Letters
Volume	23
Issue number	5
DOIs	https://doi.org/10.1109/LSP.2016.2551468
Publication status	Published - 6 Apr 2016

Bibliographical note

© 2016 IEEE. This is an author-produced version of the published paper. Uploaded in accordance with the publisher’s self-archiving policy. Further copying may not be permitted; contact the publisher for details

Access to Document

10.1109/LSP.2016.2551468

Box_constr_AF_rev.pdf
© 2016 IEEE. This is an author-produced version of the published paper. Uploaded in accordance with the publisher’s self-archiving policy. Further copying may not be permitted; contact the publisher for details
Accepted author manuscript, 192 KB

Cite this

@article{cd9f630368c6475e9fd508f8c50d5134,

title = "RLS adaptive filter with inequality constraints",

abstract = "In practical implementations of estimation algorithms,designers usually have information about the range inwhich the unknown variables must lie, either due to physicalconstraints (such as power always being nonnegative) or due tohardware constraints (such as in implementations using fixedpointarithmetic). In this paper we propose a fast (that is, whosecomplexity grows linearly with the filter length) version of thedichotomous coordinate descent recursive least-squares adaptivefilter which can incorporate constraints on the variables. Theconstraints can be in the form of lower and upper bounds on eachentry of the filter, or norm bounds. We compare the proposedalgorithm with the recently proposed normalized non-negativeleast mean squares (LMS) and projected-gradient normalizedLMS filters, which also include inequality constraints in thevariables.",

author = "Vitor Nascimento and Yuriy Zakharov",

note = "{\textcopyright} 2016 IEEE. This is an author-produced version of the published paper. Uploaded in accordance with the publisher{\textquoteright}s self-archiving policy. Further copying may not be permitted; contact the publisher for details",

year = "2016",

month = apr,

day = "6",

doi = "10.1109/LSP.2016.2551468",

language = "English",

volume = "23",

pages = "752--756",

journal = "IEEE Signal Processing Letters",

issn = "1070-9908",

publisher = "IEEE",

number = "5",

}

TY - JOUR

T1 - RLS adaptive filter with inequality constraints

AU - Nascimento, Vitor

AU - Zakharov, Yuriy

N1 - © 2016 IEEE. This is an author-produced version of the published paper. Uploaded in accordance with the publisher’s self-archiving policy. Further copying may not be permitted; contact the publisher for details

PY - 2016/4/6

Y1 - 2016/4/6

N2 - In practical implementations of estimation algorithms,designers usually have information about the range inwhich the unknown variables must lie, either due to physicalconstraints (such as power always being nonnegative) or due tohardware constraints (such as in implementations using fixedpointarithmetic). In this paper we propose a fast (that is, whosecomplexity grows linearly with the filter length) version of thedichotomous coordinate descent recursive least-squares adaptivefilter which can incorporate constraints on the variables. Theconstraints can be in the form of lower and upper bounds on eachentry of the filter, or norm bounds. We compare the proposedalgorithm with the recently proposed normalized non-negativeleast mean squares (LMS) and projected-gradient normalizedLMS filters, which also include inequality constraints in thevariables.

AB - In practical implementations of estimation algorithms,designers usually have information about the range inwhich the unknown variables must lie, either due to physicalconstraints (such as power always being nonnegative) or due tohardware constraints (such as in implementations using fixedpointarithmetic). In this paper we propose a fast (that is, whosecomplexity grows linearly with the filter length) version of thedichotomous coordinate descent recursive least-squares adaptivefilter which can incorporate constraints on the variables. Theconstraints can be in the form of lower and upper bounds on eachentry of the filter, or norm bounds. We compare the proposedalgorithm with the recently proposed normalized non-negativeleast mean squares (LMS) and projected-gradient normalizedLMS filters, which also include inequality constraints in thevariables.

U2 - 10.1109/LSP.2016.2551468

DO - 10.1109/LSP.2016.2551468

M3 - Article

SN - 1070-9908

VL - 23

SP - 752

EP - 756

JO - IEEE Signal Processing Letters

JF - IEEE Signal Processing Letters

IS - 5

ER -