Predictive quantization of spectral amplitudes for harmonic coders

Author

Prased, R. ; Chan, Wai-Yip

Author_Institution

BBN Technol., Cambridge, MA, USA

fYear

2000

fDate

2000

Firstpage

47

Lastpage

49

Abstract

We present a novel predictive vector quantization scheme for coding the variable-dimension spectral amplitude vectors produced by harmonic coders. The scheme has a safety-net prediction structure, but it uses analysis-by-synthesis codebook search. A “closed-loop” codebook design algorithm is devised. Significant improvement over conventional predictive VQ design methods is obtained

Keywords

harmonic analysis; prediction theory; spectral analysis; speech coding; speech synthesis; vector quantisation; analysis-by-synthesis codebook search; cepstral parameters; closed-loop codebook design algorithm; discrete all pole model; discrete cepstrum modelling; harmonic coders; harmonic speech coding; predictive VQ; predictive vector quantization; safety-net prediction structure; spectral amplitudes; speech spectrum representation; variable-dimension spectral amplitude vectors; Algorithm design and analysis; Cepstral analysis; Cepstrum; Design methodology; Digital audio players; Frequency; Predictive models; Sampling methods; Speech coding; Vector quantization;

fLanguage

English

Publisher

ieee

Conference_Titel

Speech Coding, 2000. Proceedings. 2000 IEEE Workshop on

Conference_Location

Delavan, WI

Print_ISBN

0-7803-6416-3

Type

conf

DOI

10.1109/SCFT.2000.878390

Filename

878390