A method for vivo detection using gesture recognition

Abstract

The invention discloses a method for vivo detection by using gesture recognition, which is characterized in that the method comprises a training data acquisition unit, a cascade analyzer unit, a facedetection unit and a living body inference reliability unit. The training data acquisition unit collects a plurality of gesture actions, and gives each gesture action command to save and create a corresponding folder in a jpg picture format; the cascade parser unit trains the collected actions, trains each folder in the training data acquisition unit, and generates a cascade parser after each folder is trained. The face detection unit is used for detecting the face of the recognition object, and then recognizing the instruction gesture. The living credibility unit scores the instruction gesture made by the recognition object, and then scores according to the recognition result of the gesture to obtain the living credibility data and the conclusion. The invention effectively avoids photo and video deception, loss of targets and other shortcomings.

Description

technical field [0001] The invention relates to a method for detecting a living body by using gesture recognition, and belongs to the technical field of face recognition systems. Background technique [0002] With the rapid development of artificial intelligence technology, its application fields are also expanding. As one of the important technologies in the field of artificial intelligence, face recognition has quickly entered people's lives, such as mobile payment, identity verification, and intelligent robots with face recognition. etc. Every face recognition system will face fraudulent behaviors such as photos, videos, and models when performing face comparisons. Especially in the application of the financial field, liveness detection is a major obstacle for face recognition technology to move to a higher level. [0003] At present, the widely used living detection methods mainly include living detection based on biological characteristics, living detection based on co...

AI Technical Summary

Problems solved by technology

[0004] However, the above-mentioned liveness detection methods still have shortcomings such as the possibility of being attacked and poor experience, which can be summarized as the following points: (1) The user experience is poor, and the liveness detection based on the coordination of commands and actions requires the detected object to make accurate responses. Instructions, which often require more users, not only need to make command actions, but also need to pass action detection, which may take a long time. If the user turns left too fast, but the system does not capture the left turn, the user needs to cooperate again; blink detection, because the blink movement is relatively small and fast, it is difficult to capture, and multiple detections may be required. Both will cause users to be irritable and have poor experience; (2) The equipment requirements are high and there are many limitations; biometric-based liveness detection may need to be equipped with an infrared camera to achieve, such as iris detection, ordinary cameras cannot meet the requirements; The live detection of special equipment requires the installation of a depth camera. These equipment requirements cannot be met in fields such as mobile payment; (3) there are security risks and are easy to be deceived; live detection based on physiological characteristics may be deceived by other living bodies. For example, when performing fingerprint detection, it is replaced by another living body that is not the subject of detection. The living body detection based on the coordination of command actions may be deceived by photos, videos, etc., such as a photo of a certain action, and a series of action videos; Liveness detection based on special equipment may be deceived by the face model. For example, when using a depth camera for detection, a 3D printed head model may be regarded as a living body