Recovery of broadband speech from narrowband speech by radial basis function neural networks

This paper presents a novel method to recovery of broadband speech from narrowband speech. The approach is based on estimating independently the spectral envelope using a radial basis function (RBF) neural network, a well-known model of artificial neural networks and the excitation function. The simulation results show that the proposed method is effective in the estimation of missing frequency components. Moreover, in listening tests, around 84% of the mean opinion score (MOS) was obtained by implementing the proposed recovery scheme.