摘要
设计一种表情识别系统,采用多种采样方式和不同尺度的局部Gabor滤波器,通过主成分分析与线性判别分析对人脸表情识别系统进行特征优化选择。该系统大幅缩减特征提取及分类的时空需求量,表情识别率也有所提高。对原始图像沿垂直方向采样识别效果说明人脸垂直方向包含更多的表情信息。实验测试结果表明,Gabor变换后的人脸表情主要特征信息在不同的尺度和方向上具有集中性和冗余性,小尺度全方向的滤波器组能获得更好的识别性。
This paper investigates a facial expression recognition system based on variant sampling method and different scales of local Gabor features optimized by Principal Component Analysis（PCA）＋ Linear Discriminant Analysis（LDA）.The sampling method not only reduces the need of compute time and storage memory,but also improves the recognition rates.The result obtained from the sampling in the vertical direction expresses that this direction includes much more facial expression information.Also the influence on facial expression recognition rates based on variant Gabor filters in different scales and directions can be concluded that the primitive information of facial expression features have redundancy in scales and directions.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第18期195-197,共3页
Computer Engineering
基金
国家自然科学基金资助项目(60872084)
关键词
GABOR变换
特征提取
下采样
主成分分析
线性判别分析
Gabor transform
feature extraction
downsampling
Principal Component Analysis（PCA）
Linear Discriminant Analysis（LDA）
