Free handwritten string recognition using CNN and DP matching
Yuta Okajima, Kento Morita, Tetsushi Wakabayashi · 2022 Joint 12th International Conference on Soft Computing and Intelligent Systems and 23rd International Symposium on Advanced Intelligent Systems (SCIS&ISIS) · 2022
While handwritten zip code recognition and ledger sheet recognition are in practical use, character recognition technology for free handwritten documents is still in the process of commercialization. If character recognition for free handwritten documents is realized, scanned images of notes and memos written on paper can be converted into text data on a computer, which will be useful for keyword searches and natural language processing. One of the reasons why character recognition for free handwritten documents is still in the research stage is that it is difficult to detect lines and extract characters from a document. In order to improve the accuracy of character recognition for free handwritten documents, it is first necessary to improve the accuracy of line detection and character segmentation. In this study, we propose a method to segment characters from words or strings using CNN and dynamic programming so that the sum of character similarities is optimal, and aim to improve the accuracy of character segmentation