Recognition of handwritten Chinese characters by modified Hough transform techniques

Chinese characters are mapped from the spatial domain into the parametric one for stroke extraction, and the dynamic programming matching (DP matching) algorithm is applied to recognize Chinese characters. This method overcomes not only the problem of noise sensitivity in the local feature approach, but also the problem of being time consuming in the global feature approach. Results are presented for two experiments conducted for a database called ETL8, which contains 881 Chinese characters and 160 variations for each one, to prove the usefulness of the MHT and DP matching methods. An actual recognition rate of 94.5% was obtained for 351 Chinese characters in the ETL8.<
>