Multi-layer fuzzy HMM for online handwriting shape recognition
Cuiyun Li, Hongbing Ji, Jihong Pei · 2005
This paper discusses a novel type of fuzzy hidden Markov model (FHMM) based on a multilayer decision tree and presents its application to online shape recognition. A local feature vector is obtained by calculating a shape's absolute angle, which is used as the feature for FHMM training and recognition. The global features of the shape are incorporated into the decision tree. The multilayer FHMM can decrease the computation time in training because of the fuzziness of the model. Due to the reduction of the shape searching space by the decision tree, the recognition time is saved and the recognition rate is improved.