BANDWIDTH EXTENSION OF CELLULAR PHONE SPEECH BASED ON MAXIMUM LIKELIHOOD ESTIMATION WITH GMM
Wataru Fujitsuru, Hidehiko Sekimoto, Tomoki Toda, Hiroshi Saruwatari, Kiyohiro Shikano · Institutional Repositories DataBase (IRDB) · 2008
Bandwidth extension is a useful technique for reconstructing wideband speech from only narrowband speech. As a typical conventional method, a bandwidth extension algorithm based on minimum mean square eηor (MMSE) with a Gaussian mixture model (GMM) has been proposed. Although the MMSE-based method has reasonably high conversion-accuracy, there still remain some problems to be solved: 1) inappropriate spectral movements are caused by ignoring a correlation between frames, and 2) the converted spectra are excessively smoothed by the statistical modeling. In order to address these problems, we propose a bandwidth extension algorithm based on maximum likelihood estimation (MLE) considering dynamic features and the global variance (GV) with a GMM. A result of a subjective test demonstrates that the proposed algorithm outperforms the conventional MMSE-based algorithm.