欢迎访问《图学学报》 分享到:

图学学报

• 智能设计与数字化设计 • 上一篇    下一篇

基于汉字字体结构认知计算的多意象预测模型

  

  1. (兰州理工大学设计艺术学院,甘肃 兰州 730050)
  • 出版日期:2019-10-31 发布日期:2019-11-06
  • 基金资助:
    国家自然科学基金项目(51465037,51705226);甘肃省自然科学基金项目(2017gs10786)

Multi-Image Prediction Model Based on Cognitive Computing of  Chinese Font Structure

  1. (School of Design Art, Lanzhou University of Technology, Lanzhou Gansu 730050, China)
  • Online:2019-10-31 Published:2019-11-06

摘要: 为了揭示汉字字体与受众的情感意象之间的内在关系,从认知计算的角度出发, 探索构建一种“设计特征-结构指标-意象”的灰箱关联模型,以其预测汉字字体的多个意象。首 先依据认知计算的原理将字体结构规则抽象为知识,运用产生式规则将字体结构知识进行定量 描述,提出字重、重心、字面、字怀 4 个字体结构指标的认知计算公式,将无序的形态信息转 化为结构化的有序信息。然后基于汉字字体意象认知系统的非线性耦合的特点,发展出一种运 用多输出最小二乘支持向量回归机(MLS-SVR)进行汉字字体多意象预测的方法。将该方法对汉 字字体的 3 个意象进行预测,实验结果表明其具有良好的预测效果和精度。该模型可作为字体 智能设计系统的适应度函数,为发展字体智能设计提供有益的参考。

关键词: 汉字字体, 字体结构指标, 多意象, 认知计算, MLS-SVR

Abstract: In order to uncover the intrinsic relationship between Chinese fonts and the emotional image of audience, this paper attempts to establish a grey box correlation model of design features-structure indexes-images to calculate multiple images of Chinese fonts from the perspective of cognitive psychology. Firstly, based on the theory of cognitive computing, the font structure rules were abstracted into knowledge. The production rules were applied to quantitatively describe the font structure knowledge, and the cognitive calculation formulas of four font structure indexes were proposed, namely, font weight, center of gravity, font circumscribed polygon, and font blank, which transform disordered morphological information into structured ordered information. Then based on the nonlinear coupling system characteristics of Chinese fonts image cognition, a multi-image prediction method of Chinese fonts was developed using multi-output least squares support vector regression machine (MLS-SVR). A method for multi-image prediction of Chinese fonts using MLS-SVR was developed to predict three images of Chinese fonts. The experimental results show that it is characteristic of good prediction bility and accuracy. The model can serve as the fitness function of the font intelligent design system, and provide a useful reference for the development of font intelligent design.

Key words: Chinese fonts, font structure indexes, multi-image, cognitive computing, MLS-SVR