Speech Bandwidth Extension Using DWT-FFT-Based Data Hiding

Abstract

A novel transform-domain speech bandwidth extension algorithm is proposed to transmit information about the missing speech frequencies over a hidden channel, i.e., the related encoded spectral envelope parameters are hidden within the narrowband speech signal using discrete wavelet transform-fast Fourier transform-based data hiding (DWTFFTBDH) technique. The hidden information is recovered reliably at the receiver to produce a wideband speech signal of much higher quality. Obtained results confirm the excellent reconstructed wideband speech quality of the proposed algorithm over traditional methods.