Information Technology Reference
In-Depth Information
TaBLe 2.2
ITU P.800.1 Terminology on Telephone Transmission Quality
Methodology
Listening-only Conditions Tested
Conversational Conditions Tested
Subjective
MOS
LQS
: P.800 Listening-only tests
MOS
CQS
: P.800 Conversational tests
Objective
MOS
LQO
: P.862 PESQ
MOS
CQO
: P.562 for PSTN, not defined
for VoIP
Estimated
MOS
LQE
: Not defined
MOS
CQE
: G.107 E-model
established by ITU for the evaluation of the telephone transmission quality
[12]. There are several recommendations for evaluating the objective conver-
sational quality of a system in absolute category rating (ACR).
1.
PESQ
(ITU P.862) is an objective measure for evaluating speech
quality based on the original and the degraded waveforms. It has
been shown to have high correlations to subjective mean-opinion-
score (MOS) results for a variety of landline, mobile, and VoIP appli-
cations. Since it only assesses the LOSQ and not the effects of delay,
it must be used in conjunction with other metrics when evaluating
conversational quality.
2. The
E-Model
(ITU G.107) was designed for estimating conversational
quality in network planning. It considers the effects of the codec,
packet losses, one-way delay, and echo. It is oversimplifying because
it assumes the independence and additivity of degradations due to
LOSQ and delay. Despite a number of extensions [13-17] that try to
address its limitations, it is difficult to extend its role beyond net-
work planning and use it for evaluating conversational quality in
actual systems.
3. The
Call Clarity Index
(ITU P.561 and P.562) was developed for esti-
mating the customer opinion of a voice communication system in a
way similar to the E-Model. Although it provides models for PSTN
systems, it does not have a user opinion model for packet-switched
networks with long delays and with nonlinear and time variant signal
processing devices, such as echo control and speech compression.
At this time, there is no single objective metric that can adequately capture
the trade-offs among the factors that affect subjective conversational quality
under all network and conversational conditions.
Subjective Measures on Conversational Quality.
A user's perception of
a speech segment mainly depends on the intelligibility of the speech heard
because the user lacks a reference to the original segment. To assess sub-
jective conversational quality, formal MOS tests (ITU P.800) [12] are usually
conducted. The method asks two subjects to complete a specific task over a
communication system, ranks the quality using an ACR, and averages the
opinion of multiple subjects.
