ScholarLens Back to search
2009•Unpublished venueRequires access

An improved 8~32kbps embedded speech codec based on ITU-T EV-VBR candidate codec

Jie Xin, Changchun Bao, Maoshen Jia, Xin Liu

Open publisher page 0 citations

Abstract

A candidate codec for ITU-T Embedded Variable Bit Rates (EV-VBR) standardization designed by Speech and Audio Signal Processing Laboratory (SASPL) of Beijing University of Technology is improved in this paper. In the improved codec, the second sub-frame's spectral parameters are quantized additionally to improve the overall quantization precision. Depth-first tree search algorithm replaces the full search and focus search in searching algebraic codebook. Transformation coding (TCX) process is redesigned and applied into the higher three layers which reduces the algorithm complexity dramatically. The objective mean opinion score (MOS) test results show that the improved codec achieves comparable speech quality as G.718 recommendation codec in most test items and has a lower algorithm delay.

About this research paper

What this paper is about

A candidate codec for ITU-T Embedded Variable Bit Rates (EV-VBR) standardization designed by Speech and Audio Signal Processing Laboratory (SASPL) of Beijing University of Technology is improved in this paper. In the improved codec, the second sub-frame's spectral parameters are quantized additionally to improve the overall quantization precision. Depth-first tree search algorithm replaces the full search and focus search in searching algebraic codebook. Transformation coding (TCX) process is redesigned and applied into the higher three layers which reduces the algorithm complexity dramatically. The objective mean opinion score (MOS) test results show that the improved codec achieves comparable speech quality as G.718 recommendation codec in most test items and has a lower algorithm delay.

Why it matters

A significance statement is not available in the OpenAlex record.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

A candidate codec for ITU-T Embedded Variable Bit Rates (EV-VBR) standardization designed by Speech and Audio Signal Processing Laboratory (SASPL) of Beijing University of Technology is improved in this paper. In the improved codec, the second sub-frame's spectral parameters are quantized additionally to improve the overall quantization precision. Depth-first tree search algorithm replaces the full search and focus search in searching algebraic codebook. Transformation coding (TCX) process is redesigned and applied into the higher three layers which reduces the algorithm complexity dramatically. The objective mean opinion score (MOS) test results show that the improved codec achieves comparable speech quality as G.718 recommendation codec in most test items and has a lower algorithm delay.

Key concepts: Codec, Adaptive Multi-Rate audio codec, Codec2, Computer science, Speech coding, Intra-frame, Mean opinion score, Speech recognition

Related papers

Related paper 1Related paper 2Related paper 3
Back to paper searchBrowse research topicsOriginal source
An improved 8~32kbps embedded speech codec based on ITU-T EV-VBR candidate codec — Research Paper | ScholarLens