基于小波变换和径向基神经网络的签名识别

The Signature-image Recognition Based on Wavelet Transform and RBF Neural Network

主要研究利用小波变换和径向基神经网络进行签名图像的分类识别。它包括不同签名图像和相似签名图像的分类识别。所提出的方法包括小波域的图像特征提取和利用径向基神经网络的模式分类。采用小波的多分辨分析方法对签名图像进行时频分析特别有效。熵和能量相关特征的概念用于小波域。径向基神经网络具有快速的收敛速度和分类能力。实验仿真证实了利用小波变换和径向基神经网络进行签名图像分类识别的有效性,且成功识别率为100%。

This paper focuses on the classification recognition of signature image by using wavelet transform and RBF neural network which contains different signature-image classification and similar signature-image classification. The proposed methods include feature extraction of image in wavelet domain and mode classification through RBF neural network. The analysis in time domain and frequency domain is especially effective when adopting multi-resolution analysis of signature image. The concept of entropy and energy concerned feature is applied to the wavelet domain. The RBF neural network owns fast speed in convergency and classification. The result of simulations proves the effectiveness of this method with the successful recognition rate of 100%.