Browse Prior Art Database

Digital Signal Reconstruction

IP.com Disclosure Number: IPCOM000039709D
Original Publication Date: 1987-Jul-01
Included in the Prior Art Database: 2005-Feb-01
Document File: 2 page(s) / 14K

Publishing Venue

IBM

Related People

Galand, C: AUTHOR [+3]

Abstract

In a voice data packet switching (VDPS) errors may occur during the transmission of a digital signal. As a result, some blocks of samples may get lost. Assume that: it is impossible to retransmit them due to the real-time constraint (for example, with speech signals). ere is a strong probability that the digital signal is a sine wave, most of the time the transmission errors will not exceed 40 ms, and the number of lost samples will be less than 10 (assuming a sampling rate of 4 ms), and due to the jitter on the delay of transmission (queues in communication lines), the received samples are enqueued in a retention buffer so as to be synthesized at a constant 20 ms block period.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 54% of the total text.

Page 1 of 2

Digital Signal Reconstruction

In a voice data packet switching (VDPS) errors may occur during the transmission of a digital signal. As a result, some blocks of samples may get lost. Assume that: it is impossible to retransmit them due to the

real-time constraint (for example, with speech

signals). ere is a strong probability that the digital signal is a sine

wave,

most of the time the transmission errors will not

exceed 40 ms, and the number of lost samples will be

less than 10 (assuming a sampling rate of 4 ms), and

due to the jitter on the delay of transmission

(queues in communication lines), the received samples

are enqueued in a retention buffer so as to be

synthesized at a constant 20 ms block period. Methods are known which take advantage of the native redundancy of the digital signal to approximate the lost packet based on the knowledge of the previously received samples. This method consists of generating the lost samples by using a second-order generator. This is possible because, in the case of voiced sounds, each sub-band signal cannot contain more than one pitch harmonic and, therefore, looks like a quasi-sinusoidal signal. So, the characteristics of this generator,
i.e, the amplitude, phase and frequency, are determined at the receiver for each sub-band and for each correctly received block of samples, and used in case of loss of the next block of samples. In order to determine the period of the digital signal, the signal is scanned to find the zero crossing. Although being very efficient, this method requires storing a large amount of data, namely, 3 blocks, to make an accurate estimation of the period in all cases. We propose hereafter an alternate method for the determination of the frequency of the generator. We assume that the signal (n) is quasi-sinusoidal with pulsation W. Therefore, between three successive samples, we have the following...