Codec Selection and Traffic Analysis for Voice over IP

1
Codec Selection and Traffic Analysis for Voice
over IP
2
INTRODUCTION

• VoIP (Voice over IP) systems has been deployed
  with network and Quality of Service (QoS)
  limitations due to the availability of the
  IP-based network infrastructure which offers a
  limited quality of service.
• The Internet can be considered here as the best
  effort IP-based network, although it needs QoS
  enhancement to convey real-time multimedia
  communications properly.
• VoIP traffic analysis is essential for the
  deployment of efficient VoIP networks that can
  cost effectively meet performance requirements
  for an optimal VoIP quality of service.

3
TWO FACTORS

• The two main variables that affect the VoIP
  quality of service (QoS) delivery, are the
  bandwidth and the voice quality.
• The bandwidth needed to be minimized while
  maintaining sufficiently good voice quality.
• Bandwidth is easily quantified, and the voice
  quality can be quantified and measured using a
  standardized ranking system called the Mean
  Opinion Score (MOS).

4
VOICE QUALITY

• MOS is a five-point scale described in ITU-T
  Recommendation P.800 Excellent-5, Good-4,
  Fair-3, Poor-2, Bad-1.
• The objective with any coding technique is to
  achieve as high a ranking on this scale as
  possible while keeping the bandwidth requirement
  relatively low.
• The basis of the MOS test is a matter of people
  listening to voice samples. ITU-T P.800 makes a
  number of recommendations regarding the selection
  of participants, the test environment,
  explanations to listeners, analysis of results,
  etc. Consequently, different MOS tests performed
  on the same coding algorithm tend to give roughly
  similar results.

5
MOS TESTS

• Because the tests are subjective, however
  variations exist. One test might yield a score of
  4.1 for a particular coding algorithm, whereas
  another test might yield a score of 3.9 for the
  same algorithm. Therefore, one should treat MOS
  values with care, and at the risk of sounding
  cynical, they should perhaps be treated with some
  skepticism, depending on who is claming a
  particular MOS for a particular codec.
• Generally and within design restrictions such as
  bandwidth, most coding schemes aim to achieve or
  approach toll quality.

6
CODECS COMPARISON

• Although the definition of this term varies, toll
  quality generally refers to a MOS of 4.0 or
  higher.
• Speech Coder Bit Rate (kbps) MOS
• G.711 64 4.3
• G.726 32 4.0
• G.723 6.3 3.8
• G.728 16 3.9
• G.729 8 4.0
• GSM Full Rate (RPE_LTP) 13 3.7

7
G.711

• G.711 is the international standard for encoding
  telephone audio. It is a pulse code modulation
  (PCM) scheme operating at an 8 kHz sample rate,
  with 8 bits per sample. According to the Nyquist
  theorem, which states that a signal must be
  sampled at twice its highest frequency component,
  G.711 can encode frequencies between 0 and 4 kHz.
• G.711 offers the highest quality under ideal
  circumstances, it does not incorporate any logic
  to deal with lost packets. In contrast G.729 does
  have the capability to accommodate a lost frame
  by interpolation from previous frames (causing
  voice errors sometimes).

8
VoIP System

• A bridge was set between two hosts of different
  IP addresses to monitor the flow of the VoIP
  traffic between them.
• Voice packets were sent from both the two hosts
  in turns using VoIP software and G.711 as a codec
  algorithm.
• Different bandwidth values were applied and the
  voice quality was monitored subject to both
  variables.

9
20 kbps Interarrivals
10
50 kbps Interarrivals
11
60 kbps Interarrivals
12
Timing Comparison
13
CONCLUSION

• The task of selecting the best voice codec or set
  of codecs to use for a given network
  implementation is a matter of balancing quality
  of service with bandwidth consumption. Given the
  range of codecs available, however, the choice
  might not be that easy.
• The delay amount within the optimum bandwidth
  range due to the encoding and decoding processes
  in the sending and receiving sides was monitored
  to be 5-7 seconds.

14
FURTHER READINGS

• 1 J. E. Flood, Telecommunications Switching,
  Traffic and Networks Prentice Hall International
  (UK), 1994.
• 2 M. Ayedemir et al., Two Tools for Traffic
  Analysis, Elsevier Computer Networks, 36, 2001,
  pp. 169-179.
• 3 R. Morris, TCP Behavior with Many Flows,
  International Conference on Network Protocols
  (ICNP97), October 1997, pp. 205-211.
• 4 D. Collins, Carrier Grade Voice over IP,
  McGrow Hill, 2001

15
Q A