网络与信息安全学报 ›› 2018, Vol. 4 ›› Issue (6): 36-44.doi: 10.11959/j.issn.2096-109x.2018044

• 论文 • 上一篇    下一篇

基于面部特征点运动的活体识别方法

王宇龙,刘开元   

  1. 广州大学计算机科学与教育软件学院,广东广州510006
  • 修回日期:2018-05-04 出版日期:2018-06-01 发布日期:2018-08-08
  • 作者简介:王宇龙(1991-),男,江苏徐州人,主要研究方向为深度学习与信息安全。|刘开元(1991-),男,安徽宣城人,主要研究方向为算法优化、深度学习与图像处理。
  • 基金资助:
    国家自然科学基金资助项目(61772007);广东省科学技术规划重点基金资助项目(2016B010124014)

Living body recognition method based on facial feature point motion

Yulong WANG,Kaiyuan LIU   

  1. School of Computer Science and Education Software,Guangzhou University,Guangzhou 510006,China
  • Revised:2018-05-04 Online:2018-06-01 Published:2018-08-08
  • Supported by:
    The National Natural Science Foundation of China(61772007);Science and Technology Planning Key Project of Guangdong Province(2016B010124014)

摘要:

提出一种应用于手机移动端基于深度学习反照片及反视频的生物识别方法,该方法利用面部动作数据集训练一个LSTM网络,通过用户配合输入一段随机动作顺序的视频,对视频预处理后提取面目特征特点得到特征点数据,最后将特征点数据放入循环神经网络判断视频是否发生了伪造攻击。实验结果表明,该方法能够有效抵御照片攻击和视频回放攻击。

关键词: 活体识别, 面部特征点, 深度学习, LSTM

Abstract:

A kind of living body recognition method was proposed,which was applied in mobile terminal and based on the deep learning.A facial movements LSTM network was trained using data sets.When users input a random sequence of video,user’s facial feature point data can be gained and whether the video forgery attacks happened will be determined by input user’s facial feature point data into circulation neural network.The test data shows that the proposed method can be protected effectively from photograph attack and video replay attack.

Key words: living body recognition, facial feature point, deep learning, LSTM

中图分类号: 

No Suggested Reading articles found!