基于Adaboost-高斯过程分类的人脸表情识别

为了弥补Ababoost分类器分类精度不够、训练耗时的缺点,利用高斯过程分类器分类精度高、计算复杂度低的优势,提出一种改进的表情识别方法.该算法将高斯过程分类(GPC)和Adaboost的人脸表情识别算法相结合,在训练二分类Adaboost时利用高斯过程分类器训练弱分类器;把这些弱分类器组合成一个总分类器,将二分类Adaboost-GPC扩展为多类分类算法.采用Gabor提取面部表情特征,由于Gabor特征提取后存在维度变高、冗余大的问题,引入二维主成分分析(2DPCA)对Gabor特征进行选择.基于Cohn-Kanade和JAFFE数据库的实验结果表明,该算法在识别正确率和速度方面的表现均较好.

Facial expression recognition based on Adaboost-Gaussian process classification

By using the Gaussian process classifier’s advantages of high classification accuracy and low computational complexity,an improved expression recognition method was proposed in order to modify the Adaboost’s disadvantage of poor classification accuracy and long time consuming.The facial expression recognition algorithm combines Gaussian process classification(GPC) with Adaboost.The algorithm uses the Gaussian process classifier as weak classifier when training Adaboost.Then these weak classifiers are combined into an overall classification,and the Adaboost is extended into a multi-class classification algorithm.Gabor wavelet transformation is used to extract facial expressional features,since the high-dimensional Gabor features are redundant;the two-dimensional principal component analysis(2DPCA) is used to select these features.Experimental results based on the Cohn-Kanade database and JAFFE database show that the accuracy and recognition speed of the algorithm are inspiring.