Chinese character skeleton automatic synthesis method and large-scale Chinese character library automatic generation method

Abstract

The invention discloses a Chinese character skeleton automatic synthesis method and a large-scale Chinese character library automatic generation method. Chinese characters are regarded as point sequences (namely writing tracks), automatic synthesis of a Chinese character skeleton is carried out by constructing a recurrent neural network model FontRNN with a monotonous attention mechanism, other Chinese characters with the same style are learned and synthesized from a small number of training samples, and automatic synthesis of the Chinese character skeleton is realized. According to the method, the defects of fuzzy synthesized Chinese characters, stroke loss, structural errors and the like in the existing Chinese character synthesis technology can be overcome, and the model can learn how to synthesize thousands of rest Chinese characters with the same style from hundreds of training samples. Therefore, the cost for manufacturing the large-scale Chinese character library can be greatlyreduced, and the method has great practical application value.

Description

technical field [0001] The invention relates to a method for automatically synthesizing Chinese character skeletons based on a cyclic neural network and an automatic generation method for a large-scale Chinese character library. The cyclic neural network is used to model a small number of Chinese character trajectories in the form of a sequence, and a large number of Chinese characters of the same style are automatically generated, which belongs to artificial Intelligence, computer graphics processing technology field. Background technique [0002] With the continuous improvement of people's living standards in our country, people's demand for personalized fonts is increasing. However, due to the diversity of Chinese characters, for example, the national standard GB18030-2000 contains 27533 Chinese characters, and even the daily-used national standard GB 2312 contains 6763 Chinese characters, and the structure of Chinese characters is often intricate, so it is necessary to u...

AI Technical Summary

Problems solved by technology

However, the model based on convolutional neural network only pays attention to the image-level features of Chinese characters, and completely ignores the important stroke order information inherent in Chinese characters, so the synthesized Chinese characters will inevitably have problems such as blurring, missing strokes, and structural errors.

Images