THE PROBLEM OF MULTI-CRITERIA CHOICE OF OPTIMAL METHOD OF VECTOR QUANTIZATION OF LINEAR PREDICTION PARAMETERS
Vector quantization (VQ) is widely used in parametric coding of speech signals. Unlike scalar quantization , vector quantization gives a considerable gain in the volume of data. A number of VQ methods is known. Basically they are all focused on the reduction of indicators such as computational complexity, memory requirements, and the distortion of spectrum. Choosing the best method of VQ on all counts is the task of a multicriterion choice and it is solved using the methods of multicriteria optimization. The article provides analysis of six methods of VQ. We consider the case when for VQ ten linear spectral roots (LSR) 24 bits/frame are extracted and multicriteria optimization problem is solved using the method of ideal point.
Authors: А. P. Boiko, D. A. Ulahovich
Direction: Informatics and Computer Technologies
Keywords: Vector quantization, speech signal, linear spectral roots, distortion of the spectrum, computational complexity, memory requirements, multicriteria optimization, method of ideal point
View full article