Pattern Recognition Strategies for Interactive Sketch Composition

The design of industrial pen-based systems for sketch composition and interpretation is still a challenging problem when dealing with complex domains. In this paper, we present a method that is based on an eager interpretation of the user strokes, i.e. on an incremental process with visual feedback to the user. We present the benefits of using such an interactive approach in order to design efficient and robust sketch recognition. We focus more specifically on the requirements in order to avoid as much as possible to disturb the user. We show how we exploit pattern recognition strategies, for instance the evaluation of adequacy measures thanks to the fuzzy set theory, the exploitation of reject options, etc., to deal with each of these difficulties. The DALI method that we present in this paper has been used to design an industrial system for the composition of electrical sketches. We show how the presented techniques are used