计算机工程与应用 ›› 2020, Vol. 56 ›› Issue (16): 97-104.DOI: 10.3778/j.issn.1002-8331.1906-0058

• 模式识别与人工智能 • 上一篇    下一篇

手写液晶体数字及识别技术

丁娜,钟宝江   

  1. 苏州大学 计算机科学与技术学院,江苏 苏州 215000
  • 出版日期:2020-08-15 发布日期:2020-08-11

Handwritten Transistor Numerals and Recognition

DING Na, ZHONG Baojiang   

  1. School of Computer Science and Technology, Soochow University, Suzhou, Jiangsu 215000, China
  • Online:2020-08-15 Published:2020-08-11

摘要:

对手写数字的识别是模式识别的一个重要研究方向。通常的手写数字风格多变,无法实现高精度的识别。为此,提出一种新颖的手写数字记录方式,称为“手写液晶体数字”,进而为其设计了一种专门的识别算法。通过多个采样窗口提取图像特征,并与各类数字的标准特征向量进行相似度计算;基于贝叶斯判决原理,依据最大后验概率完成分类;建立专门的数据集并进行测评。实验结果表明,新算法具有极高的识别率,而且识别速度很快。

关键词: 手写数字识别, 手写液晶体数字, 计算机阅卷系统, 贝叶斯分类器

Abstract:

Recognition of handwritten numerals is an important research direction in pattern recognition. Usual handwritten numbers vary seriously in writing style, and it is difficult to recognize them with a high precision. Therefore, a novel method to record handwritten numerals, called handwritten transistor numerals, is proposed. Then, a specialized recognition algorithm is developed. Firstly, image features are extracted by using a set of sampling windows, which are compared with the standard feature vectors of various number categories for computing similarity values. Then, based on Bayesian inference, the classification is completed according to the principle of maximum posterior probability. Finally, a database of handwritten transistor numerals is constructed and simulation experiments are conducted for evaluation. Experimental results show that the new algorithm has a high recognition rate and a fast recognition speed.

Key words: handwritten numeral recognition, handwritten transistor numerals, computer marking system, Bayesian classifier