Abstract:  Being an identity authentication method, signature recognition has widely used and plays an important role in the all areas and sectors of modern society. Regarding to the shortcoming of single low-dimensional features which have caused the low signature recognition rate, this paper proposes a high-dimensional statistical features based Uyghur handwritten signature recognition method. According to the demand of feature extraction, preprocessing steps such as smoothing, binarization, normalization and thinning are conducted to each signature image firstly. Then 128 dimensional local center point features and 112 dimensional ETDT features are extracted separately and they are combined to form a new high-dimensional feature. Finally, distance measurement and similarity measurement methods are used for training and recognition process. The experimental results show that the proposed method obtains much better recognition rate than earlier methods, and improves the accuracy of Uyghur handwritten signature recognition effectively.

Key words: handwritten signature, local central point feature, absolute distance, cosine distance

摘要： 签名识别作为一种身份认证方法，在现代社会的各行业各领域中普遍使用，并且发挥了重要的作用。主要针对使用单一低维签名特征进行签名识别准确率不够高的问题，提出了一种基于高维统计特征的维吾尔文手写签名识别方法。首先根据特征提取的需求，对每幅签名图像进行平滑处理、二值化、归一化和细化等预处理操作；然后提取每一幅签名的128维局部中心点特征和112维ETDT特征，将得到的两种特征组合形成新的高维特征；最后分别利用距离度量和相似性度量算法进行训练和识别。实验结果显示该算法比以前算法提取的识别结果更好，有效地提高了维吾尔文手写签名的识别率。

关键词: 手写签名, 局部中心点特征, 绝对距离, cosine距离