User´s preference based ellipse-shaped sketch recognition using genetic and winnow algorithm

Sketch recognition is a process to identify shapes from user´s sketch strokes. It is an essential component for pen-based user interface. Until now, many approaches have been proposed to achieve the recognition efficiency. However, most of them focus on improving the overall performance of the system without considering individual user´s behavior or preferences. In this paper, we present an elliptic shape detection model from user´s sketch strokes based on genetic algorithm. We then use a model based on user´s preference to choose the most appropriate shape to be either a perfect circle or an ellipse. The user model is automatically constructed using the winnow algorithm, an incremental machine learning technique, based on features related to the shape obtained from the genetic algorithm. The genetic algorithm implemented in this research uses (x₀,y₀,r₀,r₁,thetas) as a set of chromosome. Fitness function evaluates the points sampled from sketch strokes whether the shape generated is proper to the strokes or not. Recognition is done once every time the user lifts up the pointer. With this approach accuracy of 89 percent was achieved using 10-fold cross-validation method.