基于G通道心率变化的非接触人脸活体检测

被广泛应用于线上支付等相关领域的安全认证的人脸识别技术仍存在被欺骗攻击的风险。提出一种利用非接触的方式从含有心率信息最高的G通道获取含有心率变化信息的经过算法处理达到活体检测的技术。由于活体心率在短时间内不会发生突变,而欺骗攻击测得的心率会变化异常,从而心率变化可以作为活体检测的依据。分别将非接触式实时获取的心率和人脸相似度数据通过变换,来获取标准差与阈值比较,综合判断得出是否为活体。经过大量实验数据分析和现有活体检测技术比较,该算法可以准确识别真人、纸质照片、电子照片和电子视屏。多种类型的抗假体证实了该方法应用于人脸识别活体检测的高效性,并指出了发展方向以进一步改进。

NON-CONTACT FACE DETECTION BASED ON G-CHANNEL HEART RATE CHANGE

Face recognition technology, which is widely used in online payment and other related fields, still has the risk of being cheated and attacked. In this paper, we proposed a non-contact method to obtain the heart rate change information from G-channel with the highest heart rate information, which was processed by algorithm to achieve living body detection. Heart rate in living body did not mutate in a short period of time, but the heart rate measured by deception attack changed abnormally, so heart rate changes could be used as a basis for living body detection. We compared the standard deviation with the threshold value by transforming the non-contact real-time heart rate and face similarity data to judge whether it was living. After a large number of experimental data analysis and comparison with the existing in living body detection technology, our algorithm can accurately identify real people, paper photos, electronic photos and electronic television screens. Various types of anti-prosthesis prove the efficiency of this method in face recognition living body detection, and point out the direction of development for further improvement.