Browse Prior Art Database

Overflow Control in Voice Coder

IP.com Disclosure Number: IPCOM000041937D
Original Publication Date: 1984-Mar-01
Included in the Prior Art Database: 2005-Feb-03
Document File: 1 page(s) / 12K

Publishing Venue

IBM

Related People

Galand, C: AUTHOR [+2]

Abstract

When using microprocessors provided with an arithmetic logic unit (ALU) which does not support overflow control and/or saturation logic, the results of arithmetic computations, such as those involved in digital filtering operations, may overflow. Unless great care is taken to avoid overflow situations an error of the magnitude of the arithmetic dynamic may occur. The overflow situation may be avoided by an overflow control gain operation, which is easy to provide for fixed coefficient digital filtering. However, in Predictive Voice Coders, adaptive filtering operations are performed which obviously means using variable coefficient digital filters. A method is provided here to enable avoiding the above-mentioned overflow to occur while using Block Companded PCM techniques for coding the voice signal.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 62% of the total text.

Page 1 of 1

Overflow Control in Voice Coder

When using microprocessors provided with an arithmetic logic unit (ALU) which does not support overflow control and/or saturation logic, the results of arithmetic computations, such as those involved in digital filtering operations, may overflow. Unless great care is taken to avoid overflow situations an error of the magnitude of the arithmetic dynamic may occur. The overflow situation may be avoided by an overflow control gain operation, which is easy to provide for fixed coefficient digital filtering. However, in Predictive Voice Coders, adaptive filtering operations are performed which obviously means using variable coefficient digital filters. A method is provided here to enable avoiding the above-mentioned overflow to occur while using Block Companded PCM techniques for coding the voice signal. According to these techniques a scale factor "A" is determined for each block of voice signal samples, which scale factor may be equal to the largest sample within the block. The proposed overflow-controlling method consists in attenuating the input speech signal whenever an overflow risk exists, which means first detecting the risk, then acting upon the input speech samples. The scale factors are interpolated from one block of samples to the next and computed, say, ten times per block duration. In other words, each block of samples provides 10 values A(k), (k = 1,..., 10). The overflow control operation will thus be executed 10 times p...