使用稀疏约束非负矩阵分解算法的跨年龄人脸识别

摘要

人脸识别技术中除光线、姿态、表情因素外,由于年龄变化而导致的人脸形状和纹理上的变化会极大程度地影响人脸识别系统性能.对此,提出了一种使用稀疏非负矩阵分解算法来实现人脸老化模拟,然后将此方法应用于具有年龄跨度的人脸识别上,通过模拟虚拟样本来增强识别效果.实验结果表明,年龄跨度对人脸识别的确有较大的影响；当系数矩阵保持稀疏时,非负矩阵分解算法具有更强的特征提取能力；经过老化模拟增加虚拟样本后,其纹理老化效果明显地提高了跨年龄段的人脸识别的性能.%For face recognition technology, apart from lighting, gesture, and expression factors, variations in shape and texture of human faces due to aging factors also significantly affect the performance of face recognition systems. Using a sparse-constrained non-negative matrix factorization (NMF) algorithm, a facial aging simulation method based on an improved prototype was first proposed and then applied to age-span face recognition to add virtual samples and heighten the recognition rate. Experimental results show that the age span indeed has a great effect on face recognition; the NMF algorithm has stronger feature extraction ability when the coefficient matrix is sparsely constrained. Furthermore, the recognition ratio is apparently improved after adding additional virtual samples by aging simulation of face texture features.