Databases Reference
In-Depth Information
Figure 4: A shape created by PROTO type ta007
interactions are assumed. Since moving an object (hereafter called “the object
move”) is one of the most important operations in the VWDB, we studied this
move intensively. Of course, beside the object move, instance object creation and
deletion, composite object creation and deletion, and querying and updating non-
spatial attribute values of an object will have to be considered to make the language
complete, but these are left as future work. We must point out here the pioneering
work in the field of multi-modal instruction with a computer system being done
by Bolt [5], as well as the work by Tijerino et al. [6] in the area of verbal interaction
of 3-D shapes in a virtual world. However, the commands being studied are not
targeted database interactions.
The two main user modalities in a multi-modal VWDB are gestures made with
the fingers of a data glove and the user's voice. Those two modalities will be
combined so that a more complex and natural interaction can be specified. In
designing a VWDB multi-modal database language, we noticed that an interaction
initiated by a gesture or a voice instruction causes a “state change” of the target
object, which is due to the change of the relationship between the user and the
target object in the virtual world database. For example, suppose that a user starts
to move a “chair” object, as shown in Figure 5. At first, the target object, i.e., the
chair, is in the “initial” state because the relationship between the user and the
object is nonexistent at this stage. Then the user may grasp the object in order to
move it. At this point the state of the object may change from the initial state to
the “ready-to-move” state. If the user's hand starts to move to the destination,
then the state of the object is in the “in-moving” state. Now suppose that the user
releases the object after reaching the desired destination. In that case, the state of
the moved object may change to the “final” state because the move is finished.
Since the object in the final state can accept the next interaction from other users,
we may set the “final” state to equal the “initial” state via null input. Therefore
the three stages for an object move are identified as follows:
