摘要
提出了基于小波变换和学习矢量量化网络相结合的新方法进行人脸识别。小波变换具有良好的多尺度特征表达能力,能将图像的大部分能量集中到最低分辨率子图像,可以很好地对图像降维和表征人脸图像的特征。LVQ算法是在有教师状态下对竞争层进行训练的一种学习算法。LVQ网络结构简单,但却表现出比BP网络更强的有效性和鲁棒性。实验表明该方法对表情和姿态变化的人脸具有良好的分类性能和识别效率。
A method of pose-varied face recognition based on wavelet transform and LNQ network is proposed. Wavelet transform has a good ability to express the multi-resolution characteristics. The most energy of an image is mainly concentrated on its lowest resolution subimage after wavelet transtform. So it can reduce the dimension and extract the main features of an image. Learning vector quantization （LVQ） is an effective learning algorithm that trains the competitive layer under supervision. It has a simple network structure, but it is very effective and robust in face recognition. The experiment result shows that the method has very good classification capability and high recognition rate for pose-varied faces.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2006年第21期47-49,共3页
Computer Engineering
关键词
小波变换
学习矢量量化
神经网络
分类
多姿态人脸识别
Wavelet transform
Learning vector quantization（LVQ）
Neural network
Classification
Pose-varied face recognition
