Chinese handwriting recognition

Viac o knihe

Designing machines that can read handwriting like humans has been a goal for over fifty years, leading researchers to explore various approaches. While many obstacles have been overcome, some challenges remain. Chinese handwriting recognition is particularly difficult due to significant stroke variations, diverse writing styles, and a large number of confusable categories. Researchers have developed sophisticated algorithms to differentiate characters and address sample variations, resulting in substantial advancements in this field. The book introduces essential algorithms and applications in Chinese handwriting recognition. The first part covers both widely used canonical algorithms and newly developed scalable methods, systematically presenting the recognition of Chinese handwritten text. It includes guidelines for sample collection, novel recognition paradigms, distributed discriminative learning of appearance models, and contextual model estimation for large categories, highlighting celebrated methods like Gradient features, MQDF, and HMMs. The second part focuses on creating a more user-friendly human-machine interface through Chinese handwriting recognition, exemplified by four scenarios: grid-assisted input, shortest moving input, handwritten micro-blog, and instant handwriting messenger. The book progresses from basic to complex approaches and provides a reading list with literature comments.