Handwritten Chinese character recognition by ARG matching using self-organising Hopfield neural network

A model-based handwritten Chinese character recognition system is proposed. To match the attributed relational graphs (ARG) of the models with the input, the homomorphic graph matching strategy and the self-organising Hopfield network are employed. The homomorphic mapping technique is capable of interpreting an input with multiple instances of models. Hence, unsegmentably connected handwritten characters can be recognised by the proposed approach. Further, the self-organising scheme eliminates the need for specifying the constraint parameter a priori and also offers a cost effective parallel hardware implementation.