基于能量解析的计算机笔迹特征提取

Computer Handwriting Feature Extraction Based on Energy Resolution

将汉字书写过程理解为笔迹能量的空间分布过程,提出了一套有效的计算机笔迹纹理特征分析方法。通过提取分布在不同小波包最好基所对应的频率域中的笔迹能量,一幅汉字笔迹图像可以被压缩为一个含有15个元素的能量测度矢量,再由BP神经网络即可完成对经规范化后的能量测度矢量的正确分类。

Writing can be regarded as the spatial distribution process of handwriting energy. Inspired by this view, a method for analyzing handwriting texture is proposed in the paper. Through extracting handwriting energy in different frequency domains that correspond to different wavelet packet bases, a Chinese character image can be compressed into an energy measure vector of 15 elements. A BP neural networks is designed to learn and classify the results from the combination and standardization of every energy measure vector. Experimental results have confirmed the validity of the proposed method.