The Wii3D Project Conclusion
Additions
Gestures
There are several gestures that could be used in addition to those already defined.
- Double Click (used in the same way that a mouse double click is used)
- Flick (could be used for page turns, and would be especially useful for document readers)
- Scaling an Object (useful for 3D vector based programs such as Autodesk Maya or 3DStudio Max - could be defined as placing two fingers on an object and moving third towards or away from this pivot to scale the object using the third finger's direction vector)
Typing
A typing interface could be used using the methods described in [writingMarkov]. In order to create the necessary two dimensional gesture from the three dimensional paths that the Wii3D System uses, there are several techniques that could be used:
- Additions
- Double Click
- Swipe
- Scale Object
- A linear regression two decompose the three dimensional gesture into a closest fit two dimensional plane
- A removal of one of the principal dimensions (e.g. the Z plane)
- The use of a single Wii Remote's two dimensional coordinates
The linear regression would allow the user to make letter gestures in any orientation because the adjustment of the axes to become parallel to the gesture's plane and the subsequent reduction of one of the unused principal axes results in the best fit two dimensional gesture. Each letter gesture would need to be made such that it fits inside an approximate plane. The decision whether to use the gesture as a letter gesture could be ascertained by the statistical significance of the regression.
The removal of one the axes would limit a user to making gestures such that each letter gesture would fit inside the XY plane. Similarly, the use of only one of the Wii Remote's reported coordinates would limit the gesture such that its containing plane is almost perpendicular to that particular Wii Remote's line of sight.
Gesture Recognition Methods
The Wii3D System makes use of Finite State Automata and Hidden Markov Models for gesture recognition. Artificial neural networks could be implemented in the system to add another dimension to the recognition.