Minimum Error Discriminative Training for Radical-Based Online Chinese Handwriting Recognition

Free style Chinese handwriting recognition continues to pose a challenge to researchers due to the variety of Chinese writing styles. To recognize handwritten characters in an online mode, Hidden Markov Model (HMM) has been naturally adopted to model the pen trajectory of a character and a decent recognition performance is achieved. In this study, we start from a maximum likelihood trained HMM model and focus on minimizing the errors on the radical (sub-character) level to optimize the recognition performance. A novel Minimum Radical Error discriminative training criterion is proposed, and compared with the discrimination on the character level, our new approach further reduces the character error rate by 15.55% relatively (totally 29.00% reduction from the maximum likelihood baseline model) on a Chinese database,

• Type: Conference
• Year: 2007
• Where: ICDAR
• Authors: Y. Zhang, P. Liu, F. Soong
• Keywords: Chinese Handwriting Recognition, Document Analysis, ICDAR 2007, Likelihood Trained Hmm, Maximum Likelihood