基于小波分析和EMD的手写阿拉伯数字字符特征表示 |
| |
引用本文: | 李合龙;王文波.基于小波分析和EMD的手写阿拉伯数字字符特征表示[J].华南理工大学学报(自然科学版),2010,38(6). |
| |
作者姓名: | 李合龙;王文波 |
| |
作者单位: | 华南理工大学电子商务系;香港理工大学计算机系;武汉科技大学信息与计算科学系 |
| |
摘 要: | 经验模态分解(EMD)能有效地对信号结构做出精确的分辨,
利用这一特点本文提出了一种基于小波变换和EMD的票据手写体数字特征抽取方法。
通过对原始数字字符进行G小波变换极大模预处理,得到能反应字符特征信息的光滑轮廓,
进而对规范轮廓曲率序列作EMD分解,以获取浓缩曲率特征的主要信息,最后对此曲率特征数据进行聚类分析。
实验表明,与经典的字符特征提取算法相比,本文方法具有更好的聚类效果,提高了分类器的分类设计能力。
|
关 键 词: | 小波变换 经验模式分解(EMD) 曲率 特征提取 |
收稿时间: | 2009-7-14 |
修稿时间: | 2010-1-26 |
The character representation of Handwritten Arabic Numerals Based on Wavelet Analysis and EMD |
| |
Abstract: | The Empirical Mode Decompose(EMD) can
recognize effectively the structure of original signal. According to
the character of EMD, a new feature extraction algorithm of
handwritten Arabic numerals based on wavelet transform and EMD is
proposed. First, The smooth contours of numeral image are obtained
through the preprocessing of the maximum module of the G wavelet
transform. Then we apply EMD analysis to decompose the synthetic
shift normalization of curvature into their components, which could
produce more compact features. Finally, the key parameter such as
standard deviation has been selected to characterize the clustering
features analysis of arabic numerals. Experiments show the proposed algorithm can obtain better clustering
effects and improve the classifier design compared with the classic
feature extraction algorithm. |
| |
Keywords: | Wavelet transform Empirical mode decomposition (EMD) Curvature Feature extraction |
|
| 点击此处可从《华南理工大学学报(自然科学版)》浏览原始摘要信息 |