Third ESCA/COCOSDA Workshop on Speech Synthesis

November 26-29, 1998
Jenolan Caves House, Blue Mountains, NSW, Australia

On Subband Analysis Based on Glottal-ARMAX Speech Model

Keiichi Funaki, Yoshikazu Miyanaga, Koji Tochinai

Department of Electronics & Information Engineering, Graduate School of Engineering, Hokkaido University, Sapporo, Japan

We have already developed a speech analysis method based on the Glottal-ARMAX (Auto Regressive and Moving Average eXogenous) model, in which the speech production model is supposed to be an ARMAX vocal tract model and two kinds of excitation: glottal source model excitation and white Gaussian. The speech analysis method based on the Glottal-ARMAX model can estimate the glottal source and ARMAX model parameters simultaneously with pitch synchronous. In this paper, a subband processing with QMF filterbank or Haar filterbank is introduced to the Glottal-ARMAX method in order to reduce the computation. The introduction makes it possible to reduce computation since the orders of ARMAX identification can be set smaller than that of fullband analysis. The introduction also enables to improve an estimation accuracy of the glottal source model parameters owing to the improved resolution in the frequency domain.

Full Paper

Bibliographic reference.  Funaki, Keiichi / Miyanaga, Yoshikazu / Tochinai, Koji (1998): "On Subband Analysis Based on Glottal-ARMAX Speech Model", In SSW3-1998, 255-260.