公开(公告)号：GB2364870A

公开(公告)日：2002-02-06

申请号：GB0017145

申请日：2000-07-13

Applicant: MOTOROLA INC

Inventor： GIBBS JONATHAN ALASTAIR ,  JASIUK MARK A ,  EVANS ALUN CHRISTOPHER ,  SMITH AARON

IPC: G10L19/07 ,  G10L19/06

Abstract: A method of providing robust quantization of speech spectral parameters tolerant to spectral balance and speaker variations, includes the steps of, for each of a plurality of line spectral frequencies (LSFs) of a speech spectrum, quantizing (14) the displacement (12) of the LSF from an estimate (42) of its long-term mean, and, at an encoder (10) and a decoder (20), reconstructing (22,34) an estimate of the LSF from the quantized displacement and the long-term LSF mean estimate, and filtering (26,30,32) the reconstructed LSF estimate, thereby providing a subsequent long-term LSF mean estimate. In an alternative (fig 2), a prediction error, derived from the LSF from which a current short-term LSF mean value and a current moving average predicted LSF estimate have been subtracted, is quantized and, at an encoder and a decoder, the prediction error is dequantized and a next-current short-term LSF mean value and a next-current moving average predicted LSF estimate are determined.