Automatic Handwriting Verification and Suspect Identification for Chinese Characters Using Space and Frequency Domain Features

Automatic handwriting verification is to identify whether the script was written by a person himself or forged. Compared to related works about handwriting verification, the proposed algorithm adopts the features in both the time domain and the frequency domain. Moreover, in addition to distinguishing the forged manuscript from the genuine one, the proposed algorithm can also identify the suspect. The proposed algorithm is robust to writing instruments. In addition to the information of the luminance of the script, we also adopt the energy distribution on the 2-D frequency domain, the Pearson product-moment correlation coefficient (PPMCC) with genuine scripts, and vital information on characterized script points. Simulations show that the proposed method outperforms many advanced methods, including the deep-learning based method and manual identification by human beings. The proposed algorithm can well identify the script even if it is forged after several times of practice.