A Study Of ECG & PPG-based Biometrics Technology

At present,the development of user terminals is moving toward the trend of interconnection.The fundamental requirements of the user terminals are high portability,battery capacity and private informat ion security.One of the rapidest developed user terminals is wearable equipment which has quantities of users.In the past,there have been important information leakage event from user terminals,which makes information security has become one of the most important issues that wearable equipments need to be considered.The former information security mainly depends on setting the password or verification code,from past to now,there have been many biometric-based identification techologies come to application,and one of the most widely used BITs is fingerprint identification.Fingerprint identification can achieve a high accuracy,while its potential duplicability and easily change making it meets a bottleneck.Recently,there are some ECG-based identification devices,because it's difficult to replicate,there has been a wide range of attention and research on it.However,because its portability and accuracy of the acquisition device is dilemma,its application on the wearable device are blocked.In thi s paper,the data fusion and identification algorithm based on ECG and PPG signals solved this problem,because PPG signal acquisition portability and accuracy is high,which can make up for the dilemma of portability and accuracy of ECG signal acquisition device.In order to solve the problem of identification based on ECG signal,this paper proposes a data fusion method of ECG signal and PPG signal.ECG and PPG signals acquired from the wearable device are used as experimental data.The test data and the sample data are separated by one week,which verifies the stability of the algorithm for data identification in different periods.The biometric characteristics of ECG and PPG signals were analyzed in detail.Eleven time-domain features,eight frequency-domain features of the ECG signal,six time-domain features and six frequency-domain characteristics of the PPG signal were extracted.In the frequency-domain,this paper proposes a feature extraction method based on Hilbert's marginal spectrum.In the Hilbert marginal spectral of ECG signal,this paper creatively proposes a feature of extracting the decline fitting curve of the minimum radius of curvature of the eigenvalue.What's more,the unique characteristics of the Hilbert marginal spectrum of the PPG signal are also found.The amplitude distribution of the frequency blows 1 Hz and between 1-2Hz is considerably different among different individuals.In this paper,a feature fusion model based on stepwise selection and D-S evidence theory is proposed.The reliability coefficient and weight of the features are calculated by stepwise selection.Then,the D-S evidence theory is improved from the source of evidence by support vector machine theory,and according to the actual situation,this paper proposes a basic probability assignment function which is suitable for the identification experiments,and t he final identification accurancy reached 97.8%.