Off-line hand writing Chinese character segmentation method with compromised geomotric cast and sematic discrimination cost

Abstract

The geometric cost-semantic identification cost amalgamated off-line hand-written Chinese character segmentation method belongs to the field of character identification technology. Said method includes the following steps: firstly, analyzing row image of inputted off-line hand-written Chinese character, extracting stroke segment, combining stroke segments into subcharacter block, at the same time giving out subcharacter blocks combined geometric cost, utilizing these geometric costs to generate several possible candidate segmentation methods, evaluating every method by using binary syntactic model of language to obtain semantic-identification cost of every segmentation method, finally integrating geometric and semantic identification costs to obtain optimum segmentation identification scheme.

Description

technical field [0001] An off-line handwritten Chinese character segmentation method based on the fusion of geometric cost and semantic-recognition cost belongs to the field of character recognition. Background technique [0002] Optical Character Recognition (OCR) technology has always been a hot issue in pattern recognition, and Chinese character OCR technology for Chinese character recognition has a development history of more than 20 years. The recognition of offline Chinese characters refers to the recognition of Chinese character images obtained by scanners, digital cameras or cameras ( image 3 ), the recognition of offline handwritten Chinese characters has always been a difficulty. This is because people's natural writing has a relatively large degree of freedom, and it cannot provide more additional information like online handwritten Chinese characters. [0003] A key issue involved in offline Chinese character recognition is character segmentation. This is becau...

AI Technical Summary

Problems solved by technology

[0005] Chinese characters have a very rich structure, and quite a few of them have a left-right font structure. Under the writing habit from left to right, an unavoidable problem in offline Chinese character segmentation is that Chinese characters with left-right structure are often separated, such as "village" The word "village" is divided into "wood inch". For such segmentation, the recognizer often gives a very large confidence level, so the method of relying solely on the confidence level of the classifier is also unreliable.

Images