10-11-2012, 02:37 PM
Voice Quality Prediction Models and Their Application in VoIP Networks
Voice Quality Prediction Models.pdf (Size: 1.18 MB / Downloads: 42)
INTRODUCTION
IP NETWORKS are on a steep slope of innovation that will
make them the long-term carrier of all types of traffic, including
voice. However, such networks are not designed to support
real-time voice communication because of their variable
characteristics (e.g., due to delay, delay variation and packet
loss) which lead to a deterioration in voice quality [1], [2]. A
major challenge in such networks is how to measure or predict
voice quality accurately and efficiently for Quality-of-Service
(QoS) monitoring and/or control purposes to meet technical/
commercial requirements (e.g., service level agreements).
NEW METHODOLOGY FOR NONINTRUSIVE VOICE QUALITY PREDICTION
A. Introduction to New Methodology
Fig. 1 depicts a simplified, conceptual diagram of the proposed
novel methodology for developing and using new models
for nonintrusive prediction of voice quality in IP networks. The
lower part of the figure illustrates how a new model would be
used to predict end-to-end, conversational voice quality, nonintrusively,
from network and other system parameters (e.g.,
packet loss, delay and codec type). In practice, IP packets transporting
voice data through the network would be captured at a
monitoring point which may be at any suitable location (e.g., at
a gateway). Network and other relevant system parameters (e.g.,
delay, packet loss, jitter and codec type) are then extracted from
analysis of the headers (e.g., RTP headers). The parameters are
then applied to the new model to provide a prediction of voice
quality.
CONCLUSION
In this paper, we have presented a new methodology for
developing models for nonintrusive prediction of voice quality.
Based on the new methodology, we have developed nonlinear
regression models to predict perceived voice quality nonintrusively
for four modern codecs (i.e., G.729, G.723.1, AMR, and
SUN AND IFEACHOR: VOICE QUALITY PREDICTION MODELS AND THEIR APPLICATION IN VOIP NETWORKS 819
iLBC). The method exploits the intrusive algorithm, PESQ,
and a combined PESQ/E-model structure to provide a perceptually
accurate prediction of voice quality nonintrusively,
which avoids time-consuming subjective tests. We further
applied the regression models to two main applications: voice
quality prediction for real Internet VoIP traces and perceived
quality-driven playout buffer optimization. For voice quality
prediction, results show that high prediction accuracy was
obtained from the regression models (correlation coefficient of
0.987 for Scheme I and 0.985 for Scheme II, respectively) using
real Internet VoIP trace data. For playout buffer optimization,
the proposed perceptual optimized playout buffer algorithm
also achieved optimum voice quality when compared to five
other buffer algorithms for all the traces considered.