Browse Prior Art Database

A Scheme for reducing distortion of IMBE tones by mixing harmonics of a pitch

IP.com Disclosure Number: IPCOM000011397D
Original Publication Date: 2003-Feb-18
Included in the Prior Art Database: 2003-Feb-18
Document File: 2 page(s) / 31K

Publishing Venue

Motorola

Related People

Dipendra Chowdhary: AUTHOR

Abstract

IMBE (Improved Multi-Band Excitation) is a technique for compression and encoding of voice signal. In a digital communication system, a tone is often coded into IMBE vectors at the source and the IMBE vectors are decoded at the destination. This coding and decoding causes significant distortion in the tone.

This text was extracted from a Microsoft Word document.
This is the abbreviated version, containing approximately 50% of the total text.

A Scheme for reducing distortion of IMBE tones by mixing harmonics of a pitch

By Dipendra Chowdhary

 
 

Introduction                 

IMBE (Improved Multi-Band Excitation) is a technique for compression and encoding of voice signal. In a digital communication system, a tone is often coded into IMBE vectors at the source and the IMBE vectors are decoded at the destination. This coding and decoding causes significant distortion in the tone.

The author is aware of the following two methods that overcomes the problem of tone distortion. In the first method, the source sends the tone to an IMBE coder just like a voice signal. The IMBE coder detects the tone. On detection and identification of a tone, the IMBE coder stops coding and starts outputting pre-stored IMBE vectors. The coder has a set of IMBE vectors for each tone of its interest.


In the second method, the source sends a control signal instead of the audio tone to the IMBE coder. The IMBE coder uses some of the 'illegal' IMBE vectors to inform the IMBE decoder about the tone. The IMBE decoder generates the tone.

Background
A basic understanding of what happens to voice in a digital system is a prerequisite for understanding the reasons for distortion. The steps involved in coding a digitized voice (or tone) are as follows:


1. The digitized voice signal is divided into frames of equal duration (say 20 ms). The IMBE parameters are assumed to be constant over each frame.


2.Each voice frame is modeled as the response of a linear filter to some excitation signal.

3.The linear filter is a model of the envelope of the spectrum of the frame.


4.To find the parameters of the excitation signal, th...