Game Development Reference
In-Depth Information
Experimental Results
In the first part of our experiments, the efficiency of the features used to
discriminate the various gesture classes is illustrated (Figure 24 to Figure 27).
The first column shows a characteristic frame of each sequence and the tracked
centroids of the head and left and right hand, while the remaining two columns
show the evolution of the features described in the first row of Table 2, i.e., the
difference of the horizontal and vertical coordinates of the head and hand
segments. In the case of the first sequence, the gesture is easily discriminated
since the vertical position of the hand segments almost matches that of the head,
while in the closing frame of the sequence the three objects overlap. Overlapping
is crucial to indicate that two objects are in contact during some point of the
gesture, in order to separate this sequence from, e.g., the “lift of the hand”
gesture. Likewise, during clapping, the distance between the two hand segments
is zeroed periodically, with the length of the in-between time segments providing
a measure of frequency, while during the “italianate” gesture the horizontal
distance of the two hands follows a repetitive, sinusoidal pattern.
Figure 24. Hands over the head.
(a)
(b)
(c)
Figure 25. Italianate gesture.
(a)
(b)
(c)
Search WWH ::




Custom Search