Design of an adaptive FIR filter using symmetric neural networks

Tetsuya Hoya, Hiroyuki Kamata, Yoshihisa Ishida

Abstract: The LMS algorithm is generally used to design an adaptive filter. In this paper, the authors provide a new approach to design an adaptive filter using neural networks with symmetric weights trained by the modified momentum method, which is based on the backpropagation learning algorithm. The proposed method can accelerate the computation time about 15%, in comparison with the conventional LMS method.