摘要
提出一种基于模糊积分融合方式的多神经网络离线签名识别方法。首先对签名图像进行一系列预处理,提取不同类型的特征,然后使用神经网络进行初分类,再用模糊积分进行融合。实验表明,该识别方法能够比单个神经网络有更好的签名识别效果。
In this paper, an effective method based on multiple neural networks fused by fuzzy integral is presented. Firstly, a set of preprocessing is used in signature images and several different features are extracted. Secondly, these features are applied to classify elementary recognition. Finally, the results of elemtary recognition are fused by fuzzy integral. The experiment results show that the method of signature recognition is more effective than single neural network.
出处
《贵州师范大学学报(自然科学版)》
CAS
2008年第3期87-90,共4页
Journal of Guizhou Normal University:Natural Sciences
关键词
神经网络
模糊积分
签名识别
neural network
fuzzy integral
signature recognition