Pitch synchronous Fourier transform using neural networks
Munehiro Namba, Yoshihisa Ishida · 2002
In this paper, we present an overall view of the adaptive discrete Fourier transform algorithm using neural networks, and its application example for adaptive filtering. The underlying concept of the proposed method is that the continuous Fourier transform can be approximated by the discrete Fourier transform. For structuring an inverse transform system, Fourier coefficients corresponding to weight values in the network are obtained by the backpropagation algorithm to minimize the error between the output of neural networks and the signals to be analyzed. Simulation results show that our method is effective and useful for the spectral analysis of the speech signals.