基于小波包的手写体签名特征提取方法

手写体签名识别的很多特征提取方法都是基于经过复杂数据预处理和分割技术的二值图像．并且特征提取过程不可逆。因为复杂的预处理、较大的计算量和签名的连笔现象使得特征提取非常困难并对识别结果产生直接的影响。为了解决以上问题，提出了基于小波包的特征提取方法．．首先在预处理过程中对签名图像进行大小归一化：其次利用小波包对签名图像进行分解以得到签名图像在二维空间上点的集合；然后用这些二维点进行签名识别。本方法的数据预处理简单，避免了复杂分割．特征提取完全可逆。实验结果表明其具有较好的抗噪性、鲁棒性、适应性和识别率，为含噪脱机手写体签名识别提供了一种解决方案。

Research on Feature Extraction Method of Handwritten Signature Based on Wavelet Packet

Many feature extraction methods of handwritten signature recognition are based on the binary image after preprocessing and segmentation techniques and the process is irreversible. Since complex preprocessing, computing capacity,the signature with pen phenomenon,feature extraction is very difficult and make a direct impact on the effect of recognition. To solve the above problem,this paper proposes a feature extraction method based on wavelet packet. First,the size of signature image is normalized at the preprocessing. Secondly,we conducted the signature image decomposition using wavelet packet in order to get the set of points on the twodimensional space of the signatures. Then we use these points for a handwritten signature recognition. The data preprocessing of this method is simple and it avoid the complex segmentation and feature extraction is complete and reversible. Experimental results show that it has better antinoise,robustness,adaptability and recognition rate. It provides a viable solution for noisy off-line handwritten signature recognition.