软组合概率神经网络分类器人脸识别方法

概率神经网络分类器具有学习速度快、易于实现的特点,而且其输出是后验概率, 使得分类器的软组合变得容易。利用概率神经网络的这些特点,提出了软组合概率神经网络分类器人脸识别方法,该方法包括3步:(1)对人脸图像做不完全小波包分解；(2)用包含低频成分的小波子空间图像训练概率神经网络分类器；(3)用模糊积分组合训练好的分类器。将该方法与3种基于矩阵子空间的人脸识别方法在JAFFE、YALE、ORL和FERET 4个人脸数据库上进行了实验比较,结果表明,提出的方法在识别精度和CPU时间两方面均优于其他3种方法。

Soft Combination of Probabilistic Neural Network Classifiers for Face Recognition

Probabilistic neural network (PNN) classifiers have fast learning speed and can be easily implemented.The outputs of PNN are posterior probabilities which facilitate the soft combination of classifiers.We proposed a face recognition algorithm named SCPNN,which combines PNN classifiers with fuzzy integral,and makes full use of the superiori-ty of PNN and ensemble learning.The main steps of the proposed method include:the incomplete wavelet packet decomposition of face images,training PNN classifiers with wavelet subspace images which include low frequency components and combination of the trained PNN classifiers by fuzzy integral.The proposed algorithm SCPNN was compared with 3 matrix subspace algorithms on 4 face databases,which are JAFFE,YALE,ORL and FERET.The experimental results confirm that the proposed method outperforms the 3 matrix subspace algorithms in recognition accuracy and CPU time.