A new intelligent decision making system combining classical methods, evolutionary algorithms and statistical techniques for optimal digital FIR filter design and their performance evaluation


KUYU Y. Ç. , VATANSEVER F.

AEU-INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATIONS, vol.70, no.12, pp.1651-1666, 2016 (Journal Indexed in SCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 70 Issue: 12
  • Publication Date: 2016
  • Doi Number: 10.1016/j.aeue.2016.10.004
  • Title of Journal : AEU-INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATIONS
  • Page Numbers: pp.1651-1666
  • Keywords: Filter design, Evolutionary algorithm, Optimization, Artificial intelligence, OPTIMIZATION

Abstract

Filtering is one of the most important processes in electrical engineering. In digital systems, there are several methods that have been developed for filter designs. In this study, a new decision making system, which can be operated online or offline, based on statistical tests are developed for choosing the most appropriate FIR filter coefficients. For this purpose, this coefficients are optimized comparatively with nine evolutionary algorithms by using combination of some of fourteen windowing and four error functions(more than six hundred different combinations) as well as can be found via nine classical methods. As the evolutionary algorithms use random variables to achieve their results, they may not always make same design on each run. Therefore, this system is need to make a valid comparison between the algorithms employed. The key feature of proposed system is artificial intelligence,phase which sorts algorithms from best to worst under certain criteria according to chosen error function after using Kruskal-Wallis and multiple comparison tests. The proposed new approach in this intelligent decision making system, which can be also used for special, practical and educational purposes, gives the best results in between the algorithms for FIR filter design according to user requirements. (C) 2016 Elsevier GmbH. All rights reserved.