摘要
针对大数据背景下人脸识别技术存在的问题,提出一种基于压缩感知的人脸识别技术架构.系统首先利用人脸训练样本优化设计投影矩阵,然后利用优化的投影矩阵进行人脸图像的压缩感知,利用同伦算法进行快速稀疏表示分类.这样人脸识别系统一方面避免大数据传输和存储压力,另一方面可以有效保证系统识别率,实验仿真证实了研究工作的有效性.
To solve the face recognition(FR)problem existed under the background of big data,a new kind of FR technology based on compressed sensing(CS)is proposed in this paper.Firstly,the projection matrix is optimized using the face train samples.Then,the optimized projection matrix are used on the face images based on compressed sensing and the homotopy algorithm is used in the compressed sparse representation classification.With these modifications,the new FR technology can avoid large data transmission and storage pressure.On the other hand,the system recognition rate can be guaranteed.The simulation experiments show that the proposed method is valid.
出处
《浙江工业大学学报》
CAS
北大核心
2016年第4期392-398,共7页
Journal of Zhejiang University of Technology
基金
国家自然科学基金资助项目(61273195
61304124
61413262
61503339)
浙江省自然科学基金资助项目(LY13F010009
LQ14F030008)
浙江省教育厅项目(Y201430687)
关键词
人脸识别
压缩感知
投影矩阵
同伦算法
face recognition
compressed sensing
projection matrix
homotopy algorithm
