摘要
随着年龄的增长,人脸特征会发生不同程度的改变,这些改变主要包括形状特征变化和纹理特征变化,从而给人脸识别问题增加了难度。为了精确地描述年龄变化条件下人脸形状和纹理特征以提高人脸识别的精度,首先将Gabor小波运用在金字塔模型上,形成Gabor金字塔特征序列,利用均值网格对Gabor金字塔特征序列进行初步降维去噪;然后将不同样本同一方向和尺度的Gabor金字塔特征进行重组;最后构造40个并行分类器,利用直接分步线性判别分析(DF-LDA)算法进行分类识别。实验结果表明,Gabor金字塔特征序列能够提高年龄变化条件下人脸识别的精度。
Facial feature change occur in varying degrees with the increase of age. These features involve the shape features and texture features, which increase the difficulty of face recognition. In order to accurately describe the facial features with change in age to improve the accuracy of face recognition, firstly, this paper applied the innovation of Gabor wavelet on the pyramid model to structure the Gabor pyramid characteristic sequence. This paper used mean grid to descend and denoise the Gabor pyramid characteristic sequence initially, and then reconstructed pyramid characteristic sequence of different samples at the same level and direction. Finally, this paper constructed forty parallel classifiers to classify by using Direct Fractional-step Linear Discriminant Analysis (DF-LDA) algorithm. The experimental results show that the Gabor pyramid characteristic sequence can improve the accuracy of face recognition with the change in age.
出处
《计算机应用》
CSCD
北大核心
2013年第3期695-699,共5页
journal of Computer Applications
基金
国家自然科学基金资助项目(61073146)
关键词
人脸识别
GABOR小波
金字塔模型
多尺度表达
直接分布线性判别分析
face recognition
Gabor wavelet
pyramid model
muhi-scale representation
Direct Fractional-step LinearDiscriminant Analysis (DF-LDA)