A cross-language performance study of vector quantisation

Author

Parry, John J. ; Burnett, Ian S. ; Chicharo, Joe F.

Author_Institution

Dept. of Electr. & Comput. Eng., Wollongong Univ., NSW, Australia

fYear

1997

fDate

7-10 Sep 1997

Firstpage

79

Lastpage

80

Abstract

This paper investigates the performance of split vector quantisation (VQ) of the line spectral frequencies (LSFs) across a set of 10 modern languages. Spectral quantisation accounts for a significant portion of the bit allocation in low-rate speech coding. Split VQ of the LSFs can achieve transparent quantisation of the linear prediction coefficients at 24 bits/frame. The codebooks are trained on individual languages and the cross-language VQ performance was measured using spectral distortion (SD). The results show that the spectral structure of the codebook training language influences the performance of the VQ. The number of bits/frame required for transparent speech varied by as much a 2 bits across languages

Keywords

linear predictive coding; natural languages; spectral analysis; speech coding; speech processing; vector quantisation; LSF; bit allocation; codebook training language; codebooks; cross-language performance; line spectral frequencies; linear prediction coefficients; low-rate speech coding; modern languages; spectral distortion; spectral quantisation; split VQ; split vector quantisation; transparent quantisation; Euclidean distance; Linear predictive coding; Natural languages; Speech processing; Testing; Vector quantization;

fLanguage

English

Publisher

ieee

Conference_Titel

Speech Coding For Telecommunications Proceeding, 1997, 1997 IEEE Workshop on

Conference_Location

Pocono Manor, PA

Print_ISBN

0-7803-4073-6

Type

conf

DOI

10.1109/SCFT.1997.623905

Filename

623905