An Efficient Approach Combining Genetic Algorithm and Neural Networks for Eigen Value Grads Method (EGM) In Wireless Mobile Communications
Abstract
The objective of this paper is combining Genetic Algorithm and PrincipalComponent Analysis (PCA) neural network for Eigenvalue Grads Method (EGM)to estimate the number of sources in wireless mobile communications. TheEigenvalue Grads Method (EGM) is a popular method for estimation the numberof sources impinging on an array of sensors, which is a problem of great interest inwireless mobile communications. This paper proposed a new system to estimatethe number of sources by applying the output of genetic algorithm and PCA neuralnetwork with Complex Generalized Hebbian algorithm (CGHA) to EGMtechnique. In the proposed model, the initial weight and learning rate values forCGHA neural network can be selected automatically by using Genetic algorithm.The result of computer simulation for proposed system showed good response byfast converge speed for neural network , efficiency and yield the correct number ofthe sources. The important feature of new system is that, the PCA of covariancematrix are calculated based on CGHA neural network instead of determining thecovariance matrix because computation of covariance matrix is time consuming.
Keywords
Principal Component Analysis, PCA Neural Network, Complex Generalized Hebbian algorithm, CGHA, Genetic Algorithm, GA, Eigenvalue Grads Method, EGM.Metrics