Humans and animals are very good at moving. This is mainly because our brains are much better at planning movement tasks than they are at other planning tasks. This is easily demonstrable by the fact that in general, computers are much better than humans at playing chess, but if we were to place a robot in front of a chess board, it would have a much more difficult problem to play and win. This is because it is very difficult to calculate movements given information about an ambiguous and changing world. The way that the brain accomplishes this task is most likely by breaking it down into smaller movements, known as motion primitives, and overlaying these to build up a complete movement. We use statistical inference techniques to extract these building blocks of motion from natural handwriting data.
Related Publications and Presentations
- Benjamin Williams, Amos Storkey, and Marc Toussaint, “Modelling Motion Primitives and their Timing in Biologically Executed Movements”, Advances in Neural Information Processing Systems (NIPS), Vancouver, 2007.
- Benjamin Williams, Marc Toussaint, and Amos Storkey, “Extracting Motion Primitives from Natural Handwriting Data”, Proceedings of the International Conference on Artificial Neural Networks (ICANN), 2006.
- Benjamin Williams, Marc Toussaint, and Amos Storkey, “A Primitive Based Generative Model to Infer Timing Information in Unpartitioned Handwriting data”, Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI), 2006.