Adaptive beamforming by using complex-valued multi layer perceptron
Suksmono A.B.a, Hirose A.b
a Dept. of Electrical Engineering, Institut Teknologi Bandung, Indonesia
b Grad. School of Frontier Sciences, University of Tokyo, Japan
Abstract
We propose a complex-valued multilayer perception (CVMLP) neural network for adaptive beamforming. The complex-valued backpropagation algorithm (CVBPA) has been used to train the network. Experiments for a narrowband signal with multiple beam pointings and multiple nulls steering has been conducted. By using a 7-2-1 CVMLP topology and linear activation function, it is demonstrated that the beamforming by using CVMLP outperforms beamforming using complex-valued least mean square (CLMS) algorithm in terms of faster learning convergence and better interferences suppressions. © Springer-Verlag Berlin Heidelberg 2003.
Author keywords
Adaptive Beamforming,Complex-valued,Complex-valued multilayer perceptron,Learning convergence,Least mean squares,Linear activation function,Multi layer perceptron,Narrowband signal