摘要
小波分解可以大幅度降低人脸图像的维数,图像的基本信息不丢失,可以很好地表征人脸特征.用低频分量和加权高频分量分别结合PCA进行特征提取,分别计算待测试人脸与低频、高频训练人脸的欧式距离,加权计算出新的距离系数,然后利用k-近邻法分类.实验表明此方法的识别率高、训练的时间短.
Wavelet decomposition is able to reduce the dimension of face image without losing image information, so it is suitable for wavelet transform to represent face feature. Face feature based on low frequency subband is extracted through PCA, and distance of testing face to training face based on low frequency subband is computed, so does face feature based on high frequency subband. Then, the new distance function can be computed by weighed distances. Experiment shows that this method can reduce training time with high classification accuracy.
出处
《吉林化工学院学报》
CAS
2008年第1期60-62,共3页
Journal of Jilin Institute of Chemical Technology
关键词
人脸识别
加权小波分解
K-近邻法
PCA
face recognition
weighed wavelet decomposition
k-near neighbor
PCA