摘要： 为克服在图像上直接使用快速离散Curvelet变换不能完全提取有用特征信息的缺点，利用HSI颜色空间各通道互不相关的特点，结合快速离散Curvelet变换，提出一种新的视频人脸识别方法，并设计一个视频人脸识别系统以证明该方法的有效性。实验结果表明，该系统只需2 s (35帧/s)即可识别人脸，正确识别率达99.5%，对表情有一定的鲁棒性，可在

关键词: 视频人脸识别, 快速离散Curvelet变换, 最小二乘支持向量机, HSI颜色模型, 特征融合

Abstract: In order to overcome the shortcomings in extracting effective features incompletely with Fast Discrete Curvelet Transform(FDCT) on the images, this paper proposes a new video face recognition method, which takes the advantages of decorrelated HSI color space channels and FDCT. A video face recognition system is designed to prove the validity of the method. Experimental result proves that it only needs 2 seconds (35 f/s) to recognize faces, its recognition accuracy is around 99.5%, and it is robust to expressions, and it can be applied in intelligent terminal.

Key words: video face recognition, Fast Discrete Curvelet Transform(FDCT), LS-SVM, HSI color model, feature fusing

中图分类号: 

• TN911.73