List of Tables

• Packets of information arriving from the glove.
• Some important statistics on the signs used.
• Results of learning algorithms with distance, energy and time features.
• Results of learning algorithms with bounding boxes
• Results of learning algorithms on positional histograms.
• Results of learning algorithms on rotation and finger histograms.
• Results of learning algorithms on histograms of and histograms.
• Results of learning algorithms on octant counts.
• Results of learning algorithms on simple time division features.
• Results of learning algorithms synthesised time division features.
• Results of learning algorithms on all attributes combined
• Results of learning algorithms on the most effective attributes combined
• Results of learning algorithms with inter-signer learning.
• Results of learning algorithms with inter-signer learning, with partial training from that person.
• Results of learning algorithms on Vamplew's handshape data
• Results of learning algorithms on Vamplew's handshape data with additional attributes.
• Results of neural nets on Vamplew's handshape data with additional attributes.
• Results of comparison between PowerGlove and CyberGlove on handshape recognition
• Results of CyberGlove data for each of the data-sets

In the Name of the One True God, the Beneficent, the Merciful

Acknowledgements

I first of all would like to thank Allah, the Creator, for giving me the ability to complete this thesis, and for guiding me while doing it.

My thanks go to my family, who have supported me, and especially my sister, Hediah, for putting up with my cooking, and for her wonderful spaghetti.

Also to the wonderful people in the Deaf community, who have put up with my bugging them, but still teaching and helping me a great deal. In particular, David Parker and Glenn Welldon at Silent Messenger, Lynne Rogers and Michelle Macguire for teaching me the little Auslan that I know, and Anne Horwood for putting me in touch with people. Especially, I would like to thank Andrew Young, who volunteered his time to provide me with some sample signs.

Academically, I would like to thank Adam Schembri for helping me with some of the technical aspects of Auslan, for helping me correct some errors I had made in the selection of signs and for providing me with sample signs. Also, Peter Vamplew for providing the handshape datasets and for offering advice. Similarly my thanks to all the people who helped with this research through the 'net -- Chris Hand, Thad Starner, David Sturman, James Kramer, Dean Rubine, Diego Montefusco and Toni Emerson among them.

Fifthly, to the staff of the School of Computer Science and Engineering for trusting me with a thesis that is out of their usual domain of study and supporting me with the resources I needed to complete this thesis. In particular to Andrew Taylor, my supervisor, to Charles Willock for giving me ideas on how to look at the data and to Phil Preston for dealing with the equipment. Also to Simon Warfield, for his help with the implementation of the IBL algorithm; to Paul Wong for his suggestions and discussion, and to Ross Quinlan for his advice on using IBL and his suggestions for further work.

Finally, in every body of work, there are innumerable people who contribute, but are forgotten inadvertently, and for this I apologise. This is because of a shortcoming on my part and not on yours.