Wideband Re-synthesis of Narrowband Speech Using Discriminative Piecewise Linear Transformation

Abstract

Artificial Bandwidth Extension (ABE) is a technique that attempts to regenerate the wideband signal (0-8 kHz) from a narrowband signal (300-3.4 kHz) in order to improve the speech quality of today's telephone systems. By employing the well-known source-filter model, ABE can be divided into 2 sub-problems, namely extension of the excitation signal and extension of the spectral envelope. Since the extension of spectral envelope has a more significant effect on perceived speech quality as compared to that of the excitation signal [1], more efforts have been put into this problem.
Several approaches to the problem of extending the spectral envelope have already been proposed. Although speech quality is improved as compared to the narrowband speech, a clear gap between the regenerated speech and original wideband speech could still be seen. In this paper, we propose a new approach to this problem by introducing a discriminative model into SPLICE [2], and describe its performance.