Chemical Symbol Feature Set for Handwritten Chemical Symbol Recognition

Abstract

There are two main types of approaches for handwritten chemical symbol recognition: image-based approaches and trajectorybased approaches. The current image-based approaches consider mainly the geometrical and statistical information from the captured images of users' handwritten strokes, while the current trajectory-based recognition approaches only extract temporal symbol features on users' writing styles. To recognize chemical symbols accurately, however, it is important to identify an effective set of important chemical features by considering the writer dependent features, writer independent features as well as context environment features. In this paper, we propose a novel CF44 chemical feature set based on the trajectory-based recognition approach. The performance of the proposed chemical features is also evaluated with promising results using a chemical formula recognition system.