Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Adaptive Audio Playout Algorithm for Shared Packet Networks

IP.com Disclosure Number: IPCOM000104378D
Original Publication Date: 1993-Apr-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 4 page(s) / 146K

Publishing Venue

IBM

Related People

Aldred, BK: AUTHOR [+3]

Abstract

Disclosed is an algorithm for the adaptive playout of digitised voice which has been communicated through a shared packet-switched network. The standard delay-smoothing algorithm is made adaptive, with parameters driven by packet arrival time and packet loss. A variable de-sampling rate is used to fit a constant length packet to an increased or decreased time interval. The algorithm ensures that voice delay is kept to absolute minimum whilst maintaining voice quality at an acceptable level. Inventive aspects are the use of the number of lost packets to drive the voice playout algorithm delay; use of average and worst case delay over a sampling interval to drive the algorithm delay together with variation in desampling rate to match a fixed size packet to a variable time interval.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 41% of the total text.

Adaptive Audio Playout Algorithm for Shared Packet Networks

      Disclosed is an algorithm for the adaptive playout of digitised
voice which has been communicated through a shared packet-switched
network.  The standard delay-smoothing algorithm is made adaptive,
with parameters driven by packet arrival time and packet loss.   A
variable de-sampling rate is used to fit a constant length packet to
an increased or decreased time interval.  The algorithm ensures that
voice delay is kept to absolute minimum whilst maintaining voice
quality at an acceptable level.  Inventive aspects are the use of the
number of lost packets to drive the voice playout algorithm delay;
use of average and worst case delay over a sampling interval to drive
the algorithm delay together with variation in desampling rate to
match a fixed size packet to a variable time interval.

      One type of data which is communicated between person to person
(P2P) users is voice, otherwise referred to as speech or audio.
Although the bandwidth requirements for voice is an order of
magnitude less than that of full motion video, voice is more
sensitive to packet loss and delay, there being no voice equivalent
to the video 'freeze frame'.    The process of communicating voice
between two P2P users consists of five stages which can introduce
fixed and variable delays into the pipeline as follows.

o   Packetization (fixed delay) - The length of the blocks, in units
    of time, into which the incoming continuous rate voice is chopped
    in order to transport it through the system.

o   Transfer to the network interface (fixed plus variable delay) -
    The time to move a voice packet from the voice input adapter
    through the workstation software and bus to the network
    interface.  This delay is mainly fixed (software code paths) but
    may have a variable component if other higher priority work on
    the workstation needs to be done.

o   Wait before being launched into the network (variable delay) - In
    the case of Token Ring, a voice workstation cannot transmit a
    packet until it receives a free token.  This is a variable time,
    the length of which depends upon network load.  Note that the
    minimum waiting time could be zero.

o   Data transfer time through the networkk (fixed plus variable
    delay) - For transfer of data between workstations and on the
    same LAN, this will be a fixed value, being simply the time to
    send the bits in the fixed size voice packet at the network bit
    rate.  If Bridges, Routers and/or Gateways are involved in the
    transfer, variability will be introduced into this delay.

o   Transfer from the network interface (fixed plus variable delay) -
    The time to move a voice packet from the network interface
    through the workstation software and bus to the voice adapter
    where the data within the voice packets will be converted back to
 ...