Improving the Codebook Design for Vector Quantization
Y. J. Liu
Abstract
Y. J. Liu
Abstract
In the vector quantization of speech signals. the codebook design is an important element. There are two major steps in the codebook design: codebook initialization and iterative optimization. The codebook initialization is especially important in order to design a better codebook. The present initialization procedures involve the random generation of initial codewords. In this paper. a non-iterative optimization procedure is used to reduce the guesswork to a minimum. Because better initial codewords are generated in this procedure, there is no empty cell problem in our design of the codebook. In comparison with a cell-splitting procedure, the S/N performance of this procedure is better by about 2 db.
OpenAlex reports 5 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In the vector quantization of speech signals. the codebook design is an important element. There are two major steps in the codebook design: codebook initialization and iterative optimization. The codebook initialization is especially important in order to design a better codebook. The present initialization procedures involve the random generation of initial codewords. In this paper. a non-iterative optimization procedure is used to reduce the guesswork to a minimum. Because better initial codewords are generated in this procedure, there is no empty cell problem in our design of the codebook. In comparison with a cell-splitting procedure, the S/N performance of this procedure is better by about 2 db.
Key concepts: Codebook, Linde–Buzo–Gray algorithm, Initialization, Vector quantization, Computer science, Quantization (signal processing), Algorithm, Iterative method