LPQ与NMF特征融合的人脸识别

Face Recognition Based on LPQ and NMF

提出一种融合局部相位量化(LPQ)和非负矩阵分解(NMF)进行人脸识别的方法.该方法首先采用LPQ算子提取分块人脸图像的LPQ直方图序列(LPQHS),根据每块的贡献度,得到权重的直方图序列(Weight LPQHS),然后采用NMF方法提取其非负子空间及其系数矩阵,最后根据最近邻原则进行识别.在AR和YALE标准人脸数据库上的实验结果表明,该方法具有较高的识别率.

A method of face recognition based on local phase quantization（LPQ） and non-negative matrix factorization（NMF） was proposed.Firstly,LPQ operator was used to extract the LPQ Histogram Sequence（LPQHS） from block face images.According to the contribution of each face block,weight LPQ Histogram Sequence（Weight LPQHS） was obtained.Secondly,NMF was applied to weight LPQHS for extracting non-negative subspace and the corresponding coefficient matrices.Finally,nearest neighbor principle was utilized in face recognition.The simulation experiments illustrated that this method had better recognition rate on the AR and YALE standard face database.