Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

A Time-Varying Analysis Based on Analytic Speech Signals

Eisuke Horita (1), Yoshikazu Miyanaga (2), Koji Tochinai (2)

(1) Faculty of Technology, Kanazawa University, Kanazawa, Japan
(2) Faculty of Engineering, Hokkaido University, Sapporo, Japan

This paper proposes an adaptive spectral estimation method for analytic speech signals. The method decreases the calculation errors of finite precision which happen in a method estimating real coefficients. A pre-processing is applied to compensate the discontinuity of a spectrum on decimated analytic signals. Furthermore, the tracking ability of the method estimating complex coefficients is discussed. Its speed is approximately as fast as a conventional method. It is shown from the results of experiments that the proposed method estimates more accurate spectra of speech signals than conventional adaptive methods estimating real coefficients.

Full Paper

Bibliographic reference.  Horita, Eisuke / Miyanaga, Yoshikazu / Tochinai, Koji (1994): "A time-varying analysis based on analytic speech signals", In ICSLP-1994, 1631-1634.