We propose an algorithm for the constrained design of FIR filters with sparse coefficients. In general filter design approaches, as the length of the filter increases, the number of multipliers used to construct the filter increases. This is a serious problem, especially in two-dimensional FIR filter designs. The FIR filter coefficients designed by the least-squares method with peak error constraint are optimal in the sense of least-squares within a given order, but not necessarily optimal in terms of constructing a filter that meets the design specification under the constraints on the number of coefficients. That is, a higher-order filter with several zero coefficients can construct a filter that meets the specification with a smaller number of multipliers. We propose a two-step approach to design constrained sparse FIR filters. Our method minimizes the number of non-zero coefficients while the frequency response of the filter that meets the design specification. It achieves better performance in terms of peak error than conventional constrained least-squares designs with the same or higher number of multipliers in both one-dimensional and two-dimensional filter designs.
Tatsuki ITASAKA
Doshisha University
Ryo MATSUOKA
The University of Kitakyushu
Masahiro OKUDA
Doshisha University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Tatsuki ITASAKA, Ryo MATSUOKA, Masahiro OKUDA, "Constrained Design of FIR Filters with Sparse Coefficients" in IEICE TRANSACTIONS on Fundamentals,
vol. E104-A, no. 11, pp. 1499-1508, November 2021, doi: 10.1587/transfun.2020KEP0014.
Abstract: We propose an algorithm for the constrained design of FIR filters with sparse coefficients. In general filter design approaches, as the length of the filter increases, the number of multipliers used to construct the filter increases. This is a serious problem, especially in two-dimensional FIR filter designs. The FIR filter coefficients designed by the least-squares method with peak error constraint are optimal in the sense of least-squares within a given order, but not necessarily optimal in terms of constructing a filter that meets the design specification under the constraints on the number of coefficients. That is, a higher-order filter with several zero coefficients can construct a filter that meets the specification with a smaller number of multipliers. We propose a two-step approach to design constrained sparse FIR filters. Our method minimizes the number of non-zero coefficients while the frequency response of the filter that meets the design specification. It achieves better performance in terms of peak error than conventional constrained least-squares designs with the same or higher number of multipliers in both one-dimensional and two-dimensional filter designs.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2020KEP0014/_p
Copy
@ARTICLE{e104-a_11_1499,
author={Tatsuki ITASAKA, Ryo MATSUOKA, Masahiro OKUDA, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Constrained Design of FIR Filters with Sparse Coefficients},
year={2021},
volume={E104-A},
number={11},
pages={1499-1508},
abstract={We propose an algorithm for the constrained design of FIR filters with sparse coefficients. In general filter design approaches, as the length of the filter increases, the number of multipliers used to construct the filter increases. This is a serious problem, especially in two-dimensional FIR filter designs. The FIR filter coefficients designed by the least-squares method with peak error constraint are optimal in the sense of least-squares within a given order, but not necessarily optimal in terms of constructing a filter that meets the design specification under the constraints on the number of coefficients. That is, a higher-order filter with several zero coefficients can construct a filter that meets the specification with a smaller number of multipliers. We propose a two-step approach to design constrained sparse FIR filters. Our method minimizes the number of non-zero coefficients while the frequency response of the filter that meets the design specification. It achieves better performance in terms of peak error than conventional constrained least-squares designs with the same or higher number of multipliers in both one-dimensional and two-dimensional filter designs.},
keywords={},
doi={10.1587/transfun.2020KEP0014},
ISSN={1745-1337},
month={November},}
Copy
TY - JOUR
TI - Constrained Design of FIR Filters with Sparse Coefficients
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 1499
EP - 1508
AU - Tatsuki ITASAKA
AU - Ryo MATSUOKA
AU - Masahiro OKUDA
PY - 2021
DO - 10.1587/transfun.2020KEP0014
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E104-A
IS - 11
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - November 2021
AB - We propose an algorithm for the constrained design of FIR filters with sparse coefficients. In general filter design approaches, as the length of the filter increases, the number of multipliers used to construct the filter increases. This is a serious problem, especially in two-dimensional FIR filter designs. The FIR filter coefficients designed by the least-squares method with peak error constraint are optimal in the sense of least-squares within a given order, but not necessarily optimal in terms of constructing a filter that meets the design specification under the constraints on the number of coefficients. That is, a higher-order filter with several zero coefficients can construct a filter that meets the specification with a smaller number of multipliers. We propose a two-step approach to design constrained sparse FIR filters. Our method minimizes the number of non-zero coefficients while the frequency response of the filter that meets the design specification. It achieves better performance in terms of peak error than conventional constrained least-squares designs with the same or higher number of multipliers in both one-dimensional and two-dimensional filter designs.
ER -