A Method for Handling the Connected Areas in Formulas

Recognition of handwritten chemical formulas is a hot area. Since the symbols are often written connected and cannot be easily segmented, the recognition of the whole formula faces challenge. In this paper, we present a method for handling the connected areas in online handwritten chemical formulas. The formula is preprocessed firstly and then both of structure features and time features are combined for segmentation. Benzene ring is a very common structure in chemical formulas, so we employ the freshman chain code to separate it from the chemical bond when they are connected. Experiment results show that the proposed method can work with a high accuracy of more than 90%.