Graphics Reference
In-Depth Information
if ( KE energy threshold ) then evaluate GestureTimeDuration dt ;
if ( dt w time threshold ) then ImpulsivityIndex = & BV /dt ;
3.2 Expressive gesture quality synthesis
In the virtual agent called Greta (Niewiadomski et al., 2011), expressive
gesture synthesis is implemented according to the model proposed
by Hartmann et al. (2005). The procedural animation generated with
Greta can be modified by the use of six high-level parameters: Overall
Activity (OAC), Repetitivity (REP), Spatial Extent (SPC), Temporal Extent
(TMP), Fluidity (FLD) and Power (PWD).
Two of them, namely the OAC and REP, infl uence the general agent's
activity. The fi rst one corresponds to the general amount of activity.
As this parameter increases (or decreases), the number of nonverbal
behaviors increases (or decreases). Each nonverbal behavior has a value
associated with it; the OAC parameter is associated to a threshold: only
gestures that overpass this threshold are displayed. The second general
parameter, REP, specifi es the possibility that the gesture stroke will be
repeated. In this manner the generation of rhythmic repetitions of the
same behavior is allowed.
The four remaining parameters SPC, TMP, FLD and PWR can be
specified for each gesture separately. SPC controls the amplitude of
movement. It implements McNeill diagram (McNeill, 1996), where
wrist positions are defined in terms of 17 sectors. SPC can be used to
create expanded vs. contracted gestures. TMP controls temporal aspect
of movement. It modifies the speed of execution of gesture phases.
Gestures are slow if the value of the parameter has a negative value,
and fast when the value is positive. It is calculated according to the
Fitts's law (Fitts, 1954) that predicts that the time required to move
to a target area in function of the target distance. Consequently the
value of temporal extent changes the duration of each phase, and
the execution time of keyframes. The third parameter, FLD, controls
the smoothness and continuity of movement (e.g., smooth, graceful
vs. sudden or jerky). Higher values allow smooth and continuous
execution of movements, while lower values create discontinuity
in the movements. Fluidity is modeled in two different manners.
First of all, its value influences the interpolation of the gesture, i.e.,
the trajectory of the movement. For this purpose it is mapped to
the continuity parameter of Tension-Continuity-Bias (TCB) splines.
Second, it also influences the transition between two gestures. The
higher is the fluidity the more likely two gestures will be fluently
co-articulated without passing through a rest intermediate position.
Search WWH ::




Custom Search