基于优选LBP与加权SVM的年龄估计

Age estimation based on selected local binary and weighted support vector machine

针对人脸识别中由于年龄变化使识别率急剧下降的问题,提出了一种基于优选局域二值模式与加权支持向量机回归相结合的年龄估计方法。该方法首先对人脸图像进行分块,提取出各分块的LBP直方图;然后采用神经网络贡献分析法计算出各个特征的贡献值,筛选掉贡献较小的特征并对筛选后的特征赋予相应的权值;最后使用加权SVM回归训练得到年龄函数估算出目标图像的年龄。实验结果表明,该方法可以较为准确快速地对人脸图像进行年龄估计。

In order to solve the problem which the rate about face recognition sharp declined due to the different age, this paper presented a new method of age estimation based on selected LBP and weighted SVM regression. In this method, divided original data into several sub-images from which extracted LBP histograms. Then calculated the contribution values of each feature by contribution analysis algorithm of neural network. After that, abandoned the features which contribute less and gave the Corresponding weights to the remained features. At last, used weighted support vector machine regression to train the vectors and gain the whole age function, so as to estimate the age of target image. Experiment results show that the method can quickly and effectively estimate the age of the human faces.