Lai ChinLun, Tai ChiuYuan
Department of Communication Engineering, Oriental Institute of Technology, New Taipei City 220, Taiwan.
Sensors (Basel). 2016 Jul 21;16(7):1136. doi: 10.3390/s16071136.
In this paper, a smart face liveness detector is proposed to prevent the biometric system from being "deceived" by the video or picture of a valid user that the counterfeiter took with a high definition handheld device (e.g., iPad with retina display). By analyzing the characteristics of the display platform and using an expert decision-making core, we can effectively detect whether a spoofing action comes from a fake face displayed in the high definition display by verifying the chromaticity regions in the captured face. That is, a live or spoof face can be distinguished precisely by the designed optical image sensor. To sum up, by the proposed method/system, a normal optical image sensor can be upgraded to a powerful version to detect the spoofing actions. The experimental results prove that the proposed detection system can achieve very high detection rate compared to the existing methods and thus be practical to implement directly in the authentication systems.
本文提出了一种智能人脸活体检测器,以防止生物识别系统被造假者用高清手持设备(如配备视网膜显示屏的iPad)拍摄的有效用户的视频或图片“欺骗”。通过分析显示平台的特性并使用专家决策核心,我们可以通过验证捕获面部中的色度区域,有效检测出欺骗行为是否来自高清显示屏中显示的假脸。也就是说,通过设计的光学图像传感器可以精确区分真实或欺骗性的面部。综上所述,通过所提出的方法/系统,可以将普通的光学图像传感器升级为强大的版本,以检测欺骗行为。实验结果证明,与现有方法相比,所提出的检测系统可以实现非常高的检测率,因此可以直接在认证系统中实际应用。