A unified framework for ACELP codebook search based on low-complexity multi-rate lattice vector quantization

Author

Zhou, Cheng ; Hu, Ruimin ; Zhang, Yong

Author_Institution

Wuhan Univ., Wuhan

fYear

2007

fDate

17-19 Oct. 2007

Firstpage

1566

Lastpage

1569

Abstract

Speech codec with ACELP structure are widely used due to their high performance for the bit rates between 4 kbps and 16 kbps. In ACELP codec, searching for the best codevector from a huge codebook is the crucial pan in both performance and complexity. In some cases, the effect of codebook search does not fit our expectation. Furthermore, there is no unified method for ACELP codebook search; rather different method is developed for each ACELP codec. In this paper, a unified framework for fixed codebook search based on Low-complexity multi-rate lattice vector quantization is proposed, and is applied to a standard ACELP codec AMR-WB+. It is confirmed that the proposed method enhances the performance especially at lower bit rates.

Keywords

speech codecs; vector quantisation; ACELP codebook search; AMR-WB+; multirate lattice vector quantization; speech codec; Adaptive filters; Decoding; Distortion; Frequency; Information filtering; Information filters; Information technology; Lattices; Nearest neighbor searches; Vector quantization; ACELP; AMR−WB+; fixed codebook search; lattice vector quantization;

fLanguage

English

Publisher

ieee

Conference_Titel

Communications and Information Technologies, 2007. ISCIT '07. International Symposium on

Conference_Location

Sydney,. NSW

Print_ISBN

978-1-4244-0976-1

Electronic_ISBN

978-1-4244-0977-8

Type

conf

DOI

10.1109/ISCIT.2007.4392266

Filename

4392266