Channel Equalization based on SFLA and DSO Trained Artificial Neural Network

Samir Saidani, et al.

Journal of Telecommunications and Radio Engineering

www.doi.org/10.1615/TelecomRadEng.v78.i17.70

Abstract

This paper presents a novel method for optimizing the weights of a neural network applied to channel equalization. Although several algorithms are found in the literature to update the weights of the neural networks, each one has their merits and drawbacks. Therefore, a learning algorithm is often modified, or even hybridized with another one in order to enhance their performances. In this context, the authors present a new approach based on two algorithms which are: Shuffled Frog-Leaping Algorithm (SFLA) and Directed Search Optimization (DSO). The updating strategy in the local search of the standard SFLA is substituted with the updating strategy of the DSO algorithm. For the global search, the authors kept the strategy of the SFLA algorithm. The proposed method is compared with the standard Particle Swarm Optimization (PSO), the Modified Shuffled Frog-Leaping (MSFLA) and the DSO algorithms to evaluate their performances for nonlinear channel equalization. The experiments reveal that the proposed method presents the best results and outperforms the other algorithms.

Keywords:

 Artificial Neural Network (ANN), Shuffled Frog-Leaping Algorithm (SFLA), Directed Search Optimization (DSO), adaptive channel equalization, digital communication