Research On Identity Authentication Based On PPG

Traditional authentication relies on user input and mainly consists of password and graphic patterns.However,these methods are known to be vulnerable to many attacks.With the development of biometric methods such as fingerprint recognition and face recognition,biometric identification is widely used with extremely high security,and authentication methods based on photoplethysmogram(PPG)have become an important authentication method.The authentication method based on physiological signals instead of face and fingerprint,because of the uniqueness of physiological signals and the characteristics of in liveness detection,its safety is very high.In this paper,A bimodal authentication embedded system based on PPG is developed.The embedded system combines biometrics and an online machine learning model.The Capnobase,BIDMC and real datasets collected by embedded system are used to validate the system accuracy with 100%,91.13% and 98.6%accuracy.The main contributions of this paper are as follows.1.A sorting fuzzy min-max neural network is proposed.Since the accuracy of the model in fuzzy min-max neural network can be affected by the input order.Before the operation of the fuzzy min-max neural network,the input data are sorted according to the smallest same class distance and the largest different class distance.The problem of input order that exists in fuzzy min-max neural network is solved,which makes the model more robust.2.A PPG-based single authentication model is proposed.The authentication model consists of three stages: signal preprocessing,feature selecting,and authentication identification.In the signal preprocessing stage,the signal noise is filtered using the Butterworth filter,the fiducial features of the signal are extracted as the original samples.In the feature selecting stage,a feature selecting algorithm based on Euclidean distance is proposed to remove outlier samples present in the original samples.In the authentication recognition stage,the fuzzy min-max neural network is used for authentication.The model has been validated by open-source dataset with good accuracy.3.A PPG-based bimodal authentication system is developed.The hardware part of bimodal authentication system consists of a controller STM32F407 and signal acquisition devices ADS1292 and Pulse Sensor.The PPG signals and ECG signals of the subject are collected by the signal acquisition module.The signals are processed and authenticated by the controller.The software part of the system combines a bimodal authentication model and a sorting fuzzy minimum maximum neural network.A feature extraction method with multi-feature fusion is also proposed,using a combination of ECG signals and PPG signals for authentication,which solves the problem of single signal dependence of the system and improves the security of the system.