Discrete cosine transform-based data hiding for speech bandwidth extension

The limited narrow frequency range of 300–3400 Hz used in public switched telephone networks causes a significant reduction of speech quality. To address this drawback, a new robust transform-domain speech bandwidth extension method is proposed in this paper. The method uses the discrete Cosine transform-based data hiding (DCTBDH) technique to provide a better-quality wideband speech signal. The spectral envelope parameters are extracted from the high-frequency components of speech signal existing above narrowband, which are then spread by using spreading sequences, and are embedded within the DCT coefficients of narrowband signal. A better-quality wideband signal is reconstructed using the extracted embedded information at the receiver end. In simulations, the high-quality wideband speech was obtained from speech transmitted over a public switched telephone network. The spectral envelope parameters of the high-frequency components of the speech signal are transparently embedded with a mean square error of 5.78 × 10^–4. In a mean opinion score (MOS) listening test, we verified that the proposed method yields improved perceptual transparency compared to conventional methods of about 0.21 points on the MOS scale. The log spectral distortion value obtained was 2.2248 which showed that the proposed technique yields an improved quality of speech signal compared to conventional methods.