基于主元分析的空间手写平面化预处理技术

Flatten preprocessing technology of 3D space handwriting based on PCA

为了将空间手写识别回归为平面手写识别问题,提出了一种新型的空间手写识别平面化预处理技术,采用基于主元分析(PCA)的投影算法对空间手写字符轨迹进行了平面化处理。该算法中,投影平面的确定仅依赖于手写字符轨迹采样点集本身的统计特征,故当书写角度发生变化时,投影平面也会随之产生适应性变化,以产生最佳的投影效果;最后在实验中,对比了指定初始平面投影法和主元分析投影法在不同书写角度下的投影效果,实验结果直观地证明了该投影算法的有效性。

In order to change 3D handwriting problem to 2D handwriting problem,a novel flatten preprocessing technology of 3D space handwriting was proposed,the locus of 3D handwriting was flatten by the algorithm based on principal component analysis（PCA）.In the algorithm,the projection plane was only determined by statistic character of the locus.The projection plane would adapt to the degree change of handwriting to generate the best projection.At last,the projecting results of both fixed plane projecting and PCA analyzed plane projecting are compared to prove the effectiveness of PCA based projecting algorithm.