Adaptive Filter Identification Using Genetic with LMS (GALMS) Algorithm

Abstract

Conventional non-adaptive filters that are used for extracting the information from an input signal, are normally linear and time-invariant. In this case, the restriction of time invariance is removed. This is done by allowing the filter to change the coefficients used in the filtering operations according to some predetermined optimization criteria. This has the important effect that the adaptive filters may be applied in areas where the exact filtering operation required may be known a priori and further, this filtering operation may be non-stationary. System modeling or system identification is one of the wide applications of adaptive filtering that have great importance in the fields of communication systems and signal processing. The main object of this paper is to find the best optimization algorithm that gives a minimum Mean Squared Error (MSE) between the desired and the actual signal to identify the unknown system. Many algorithms will be studied, such as the Least Mean Squared (LMS) algorithm, Adaptive Linear Neuron Network (ADALINE), and Genetic Algorithm (GA). Then we will produce a new improvement algorithm (we called it GALMS) that uses the LMS algorithm with an optimized learning coefficient using a genetic algorithm. Optimal weights (coefficients) will also be found to be concentrated with the actual weights.