Feasibility Research On Authentication Based On Facial Tissues Oxygen Saturation

Identification is a security assurance technology used to confirm the identity information of people.With the development of computer technology,new identification technologies based on biometric features such as face,fingerprint and vein have gradually replaced the traditional identity verification methods.However,biometric features such as faces and fingerprints are easier to be forged and spoofed,so researchers in the field of identity recognition have been working to explore new biometric features that are difficult to be forged.Facial Tissue Oxygen Saturation(StO2)reflects the oxygen-carrying and oxygen-consuming capacity of hemoglobin in the capillaries of facial tissues.It is closely related to the vascular distribution pattern and the metabolic rate of the body.Therefore,from a physiological point of view,the facial StO2 distribution pattern of each individual is unique and difficult to be falsified.However,there has been no research to explore the feasibility of identity recognition based on facial StO2.To address this issue,from the perspective of verifying the presence of identity information in facial StO2,this paper demonstrates for the first time that facial StO2 can be used for identity recognition by comparing it with face recognition in visible light.Based on this,this paper further explores the influence of stress on StO2 identity recognition.Finally,the paper proposes two methods to reduce the influence of stress changes on StO2 identity recognition.The main work and contributions of the paper are as follows.(1)Comparing the effect of identity recognition based on StO2 with that based on face images,this paper verifies that identity recognition can be done by facial StO2 alone.In the paper,a recognition model based on convolutional neural network is built,and the facial StO2 image is compared with the RGB face image in visible light for recognition without changing the model parameters and the judgment analysis method.The experimental results show that the test results of identity recognition using facial StO2 images are close to those of face recognition,which proves that facial StO2 can be used as an identity tag.(2)The influence of stress on identity information in facial StO2 is explored.Firstly,based on the publicly available StO2 stress dataset,the paper uses Res Net50 model as a feature extraction network,and uses random five-fold cross-validation to obtain StO2 identification results under multiple mixed stress states,which is then used as a baseline for comparison with subsequent experimental results.Secondly,in order to explore the identity information in facial StO2 under different states in detail,the paper sets up three experimental validation methods with baseline state training,stress state training and mixed state training respectively according to different data set partitioning methods.Finally,in order to further explore the changes of identity information in StO2 below the stress state,the paper divided the face into local regions and compared the performance of StO2 identity recognition in different RoI(Region of Interest)by the way of hypothesis testing and classification.The experimental results show that stress has a certain degree of influence on the facial StO2 distribution pattern of individuals,but the StO2 under stress state still contains abundant identity information.(3)Two methods are proposed to reduce the influence of stress on identity information in StO2.The first method is based on an ensemble classifier that first performs manual feature extraction and then combines multiple feature selection algorithms for screening.The second method designs a network module dedicated to the recognition of identity information in facial StO2.The proposed method achieves excellent results regardless of the verification method,with a maximum classification accuracy of 90.48% under stress.