Paper
Wobbrock, Jacob O., Andrew D. Wilson, and Yang Li. "Gestures without libraries, toolkits or training: a $1 recognizer for user interface prototypes." Proceedings of the 20th annual ACM symposium on User interface software and technology. ACM, 2007.
Notes
We have now finished the corner finding portion of readings and will be moving on to recognition in preparation for the next homework. The recognition portion begins with the most simple recognition algorithm, which is often used as a baseline for error evaluation. This is the $1 algorithm -- template matching. In this paper, the authors present a method for finding the optimal matching shape/template using the distance between the drawn shape and the template. There are many variants of $1, e.g. $N, $P, $3, etc., but all of them are based on the same idea of resampling into a cloud of points and finding the distances between drawn shapes and templates. Dealing with rotation differs slightly between methods, such as using polar coordinates to handle minimizing rotation, but overall, the idea is the same. Following this introduction into recognition, we'll start exploring more specific methods.
No comments:
Post a Comment