Graphics Reference
In-Depth Information
to be established (dotted links). Further, this process also merges the
repeated hold-movement phases into a single hold.
The second process ' Align CE ' then temporally aligns each
conversational expression (CE) with the input text. Each Content
unit (CU) is temporally defined by the following attributes: delay ,
duration , persistence , and duration-down . And the Phase unit (PI), stored
in the Movement Phases relation layer, is temporally defined by the
following attributes: duration-up and persistence . The temporal values
of the Content units and Phase units are calculated from the temporal
values of their children (phoneme and sil units), segment units stored
within the Segment layer (predicted by the acoustic prosody deque ). The
following equations are used:
i
= =1
CE
.delay
CE
.delay
+
CE
.duration
;
1
i
j
j
j
CE
.persisten
ce
=
t
(MP
);
2
i
n
H
= =1
n
CE
.duration
t(MP
)
CE
.persisten
ce;
3
i
i
i
i
n
=
CE
.duration
=
t(MP
)
=
t(s
)
;
4
i
down
R
i
i
k
n
=
CE
.duration
=
t(MP
)
=
t(s
)
;
5
i
up
SP
i
i
k
n
=
PI
.persisten
ce
=
t(MP
)
=
t(s
)
;
6
i
H
i
i
k
The delay attribute (equation 1) represents the absolute duration,
measured from the beginning of the input text sequence, for which
the non-verbal behavior is withheld prior to execution ( i represents
observed content item). If the first coverbal gesture does not start with
the beginning of the text sequence, a default delay is inserted, with the
duration equal to the sum of the speech fragments preceding the first
indication of a coverbal gesture. The duration of the hold-movement
phase for each CE represents the value of the persistence attribute for
the CE. The value is calculated by equation 2. This attribute represents
the duration for which the shape is maintained prior to the execution
of the next CE (if last movement, then persistence = “hold”, else
Search WWH ::




Custom Search