Browse Prior Art Database

Sub-Band Quantizing Method for Speech Signals

IP.com Disclosure Number: IPCOM000042355D
Original Publication Date: 1984-May-01
Included in the Prior Art Database: 2005-Feb-03
Document File: 1 page(s) / 12K

Publishing Venue

IBM

Related People

Galand, C: AUTHOR

Abstract

This is a method for optimally quantizing speech signals. Among the various methods used for coding speech signals are methods involving the quantizing of either signals extending over the full frequency voice bandwidth or signals extending over a limited (base band) bandwidth. In these cases either the full band or the base band is split into several sub-bands, and the signals over the various subbands are quantized by dynamically allocating the quantizing bit resources over the various sub-bands. Such a dynamic allocation is in general based on the variation of the relative energies of the signals contained in the various sub-bands. Behind all these coding methods lies the basic principle that, given a quantizing bit rate, the S/N ratio of the quantized signal may be optimized by rationally allocating the available bits.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 75% of the total text.

Page 1 of 1

Sub-Band Quantizing Method for Speech Signals

This is a method for optimally quantizing speech signals. Among the various methods used for coding speech signals are methods involving the quantizing of either signals extending over the full frequency voice bandwidth or signals extending over a limited (base band) bandwidth. In these cases either the full band or the base band is split into several sub-bands, and the signals over the various subbands are quantized by dynamically allocating the quantizing bit resources over the various sub-bands. Such a dynamic allocation is in general based on the variation of the relative energies of the signals contained in the various sub-bands. Behind all these coding methods lies the basic principle that, given a quantizing bit rate, the S/N ratio of the quantized signal may be optimized by rationally allocating the available bits. The split band technique may be further improved by conveniently processing the signal information to be quantized. The proposed method takes advantage of the short-term redundancy of the sub-band signals assuming the splitting function is performed through narrow band filtering operations. This assumption allows analyzing the sub-band signals according to a pure sine-wave model, particularly for "voiced" sounds. In other words, narrow band filtering of "voiced" sounds would lead to split signals presenting pitch and harmonic information lying within distinct sub-bands. Based on these conditions one...