融合Gabor特征与投影字典对学习的人脸识别算法

Face Recognition Methods Fusing Gabor Feature and Projective Dictionary Pair Learning

为了获得更好的人脸特征,有效地提高算法的识别率,提出了一种联合Gabor特征与投影字典对学习的人脸识别算法G-DPL。算法使用Gabor小波提取人脸图像的局部特征,对特征向量使用PCA与LDA的方法进行降维。将投影字典对学习算法与降维后的Gabor特征融合,然后进行分类识别。提出的G-DPL算法在ORL库上整体识别率达到99.00%,特征维数为39维。在AR库上识别率达到96.14%,特征维数为99维。提出的G-DPL算法在占用较少空间的同时能够获得更高的识别率,对实际应用具有一定的参考价值。

In order to obtain better face features and enhance the recognition rate of algorithm, a face recognition algorithm based on Gabor feature and projective dictionary pair learing named G-DPL is proposed in this paper. The local feature of face image are extracted by Gabor wavelet and PCA and LDA scheme is used to reduce the feature dimension. Projective dictionary pair learning algorithm and dimensionality reduced Gabor feature are fused to identify the classification. The recognition rate of G-DPL algorithm can reach 99.00% under ORL database. Feature dimensionality is 39. G-DPL can reach 96.14% on AR database. Feature dimensionality is 99. The proposed G-DPL algorithm can obtain higher recognition rate while taking up less space, which has certain reference value for practical application.