Spectral envelope estimation used for audio bandwidth extension based on RBF neural network

Abstract

In this paper a new spectral envelope estimation method based on radial basis function (RBF) neural network is proposed for implementing a blind bandwidth extension method of audio signals. To make the sub-band envelope of high-frequency (HF) components accurately recovered, the RBF neural network is utilized to fit the relationship between low-frequency (LF) features and sub-band envelope of HF components. In addition, the fine structure of HF components which can guarantee the timber of the extended audio signal is reconstructed based on nonlinear dynamics. The objective and subjective test results indicate that the proposed method outperforms the reference methods.