Information Technology Reference
In-Depth Information
MS
A
=
MED
A,B
+
HRD
B
+
MED
B,A
Person
A
A Speaks
B Heard
MED
B,A
Person
B
A Heard
B Speaks
Time
MED
A,B
MS
B
=
HRD
B
Figure 2.2
Asymmetric mutual silences.
Improved one-
way LOSQ with
increased MEDs
Improved efficiency
and symmetry with
decreased MEDs
Best MED
Mouth-to-ear delay
Figure 2.3
Trade-off considerations.
experienced by B is only governed by his or her HRD (
MS
B
=
HR D
B
). T h is asy m-
metry leads to a perception that each user is responding slowly to the other,
and consequently results in degraded efficiency and perceptual quality [4].
Conversational quality cannot be improved by simultaneously improving
LOSQ and reducing MED. A longer MED will improve LOSQ because seg-
ments will have a higher chance to be received, but will worsen the symme-
try of MSs. Figure 2.3 shows the delay-quality trade-off and a suitable MED
with the best quality. This trade-off also depends on the turn-switching fre-
quency [4,7] and on changes in network and conversational conditions [8].
It has been shown that long MEDs can cause doubletalk and interruptions
even when MED is constant [6,9].
The perceived effect of delays in multi-party VoIP are more complex than
those in two-party VoIP because there may be large disparities in network
conditions across the participants [10]. In the multi-party case, the conversa-
tional quality depends on the LOSQ and the latency of the one-way speech
from each speaker, as well as the symmetry of the conversation among the
participants. Hence, each listener may have a slightly different perception of
the same conversation. Figure 2.4 depicts two conversation units (CU) in a
