International Symposium on Chinese Spoken Language Processing (ISCSLP 2002)

Taipei, Taiwan
August 23-24, 2002

Comparison between the Spectral Estimation Techniques by Different Spectral-distortion Measures

Shaohui Zhu, Wenju Liu, Bo Xu

National Lab. of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing, China

Relative to the speech production and perception models, spectral envelopes play an important role in speech analysis, synthesis, and coding. Recently, spectral envelope estimation technique has made a rapid progress. There are several ways to obtain spectral envelope. These ways include SEEVOC technique, discrete cepstrum method, regularized discrete cepstum estimation, DAP, MVDR, etc. In this paper, we compared the different spectral estimation techniques by using the different spectral-distortion measures. It aims to compare the different envelope estimation techniques from some different perspective. The work is implemented in a low bit-rate coder based on the sinusoidal model by using the different techniques to captures the spectral amplitudes.

Full Paper

Bibliographic reference.  Zhu, Shaohui / Liu, Wenju / Xu, Bo (2002): "Comparison between the spectral estimation techniques by different spectral-distortion measures", In ISCSLP 2002, paper 41.