摘要
文章介绍一种基于主成分分析(principal component analysis,PCA)和Softmax回归模型相结合的人脸识别方法,该方法通过PCA对整幅图像提取特征,然后将提取的特征经过非线性变换输入到Softmax回归模型中。将主成分提取特征看成是单层神经网络,将它与Softmax回归模型构成的级联结构看作是2层神经网络,在神经网络的训练过程中,主成分的特征向量可以微调。在不同人脸数据库上的实验表明,相比于传统的只用PCA降维的方法,本文方法可达到较高的识别率。
In this paper, a face recognition method based on the combination of principal component analysis （PCA） and Softmax regression model is introduced. In the method, the image feature is first extracted by PCA, and then the extracted feature is input into the Softmax regression model via nonlinear transform. The PCA is considered as a single-layer neural network, so the combination of PCA and Softmax regression model can be thought as a two-layer neural network. In the training process of neural networks, the feature vectors of the principal component can be fine-tuned. The results of the experiments on the different face databases indicate that the proposed method has good recognition performance and achieves a higher recognition rate than traditional method of PCA dimension reduction.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
北大核心
2015年第6期759-763,共5页
Journal of Hefei University of Technology：Natural Science
基金
国家自然科学基金资助项目(61105076)
中国博士后科学基金面上资助项目(2012M511402)
安徽省自然科学基金资助项目(1408085MKL76)
关键词
人脸识别
主成分分析
Softmax回归模型
神经网络
face recognition
principal component analysis（PCA）
Softmax regression model
neuralnetwork
