基于离散小波变换和主成分分析的人脸识别

Face recognition based on discrete wavelet transform and principal component analysis

提出了一种基于离散小波变换、主成分分析和余弦相似度分类器的人脸识别方法。首先对图像进行二维离散小波变换,得到近似分量以及水平、垂直和对角细节分量,然后对每幅图像进行二维主成分特征提取。最后,使用余弦分类器对4种特征进行分类并融合得出最终结果。实验表明,此方法准确率要优于单一的二维主成分分析,且余弦相似度分类器拥有比欧几里得距离更好的分类效果。

A face recognition method based on discrete wavelet transform,principal component analysis and cosine similarity classifier is proposed. Firstly, the two-dimensional discrete wavelet transform is implemented to obtain the approximate component as well as the horizontal,vertical and diagonal detail components. Then the two-dimensional principal component analysis is performed on each image acquire the eigen feature. Finally,the cosine classifier gives the classification results by combining 4 kinds of features. The experiments show that the classification correctness is higher than single two-dimensional principal component analysis,and cosine similarity classifier is better than Euclidian distance classifier.