基于PIDC和二叉决策树SVM的人耳识别

Ear Recognition Based on PIDC and Binary Tree SVM Classification

提出了一种基于PIDC和二叉决策树SVM的人耳识别方法。采用PIDC方法以类间概率信息距离为监督提取人耳特征,降低了提取特征的维数;将PIDC方法与二叉决策树SVM分类方法相结合,实现了利用多类间概率信息距离监督人耳特征提取和分类。利用该方法对400个人耳进行识别实验,并将识别结果同PCA方法进行了比较,实验表明,文中方法降低了分类难度,提高了人耳识别率。

In this paper,we present an ear recognition method using PIDC and binary tree SVM classification.On the one hand,we extract ear features using PIDC which is a kind of supervised algorithm regarded for the PID between two classes.And the PIDC algorithm decreases the feature dimension under not decreasing recognition effect.On the other hand,we build a binary tree SVM classification system so that we can use Multi-class PID to guide the feature extracting process.At last,we compare the experiment results between our method and PCA using 400 ear samples,the results show that our method can reduce the difficulties of calculation and get high recognition rate.