Research Of Electrocardiogram Based Human Authentication In Different States

With the development of science and technology,people’s demand for personal information security is gradually climbing.Since biological features are a not easy to be forgotten or losing,they are applied to identity authentication in more scenes,such as payment system,immigration clearance authentication system,etc.Electrocardiogram(ECG)is a “live” signal and is very difficult to be copied or forged,which is safer to be used in authentication.Recently,more attention has been paid to ECG’s ability in authentication.Due to ECG signals are easily affected by the external environment and the states(physiological or psychological)of each individual.ECG signals of the same individual in different states are different on its shape.Therefore,finding stable features becomes one of the key issues in ECG authentication research.In this paper,13 volunteers’ ECG data were collected in different states for a period of half a year,which provided the data source for the ECG authentication in non-limited state.To find stable features of the ECG signal in different states,this paper proposes a feature extraction method based on ECG superposition number matrix by single beat ECG data.By this matrix two-dimension distribution of ECG signal are described.On the processing matrix segmentation and similarity comparison,ECG stable feature distribution area is selected,and stable feature sets are constructed.The experiment was carried out on MIT-BIH data set,self-collecting ECG data set in calm state and self-collecting ECG signal data set in different states,and the average True Positive Rate reached 90%,89.64% and 86.14% respectively.Experimental results show that with the increasing amount of ECG single beat data,the stable features of individuals are gradually revealed,the proposed method can effectively obtain stable features from ECG data in different states,and can complete authentication processing,that is to say,the authentication based on this method is effective.In the further study,it was founded that extracting stable feature distribution need a large number of one individual’s single beat ECG,by which superposition number matrix of ECG signals can be acquired.In this paper,a Matthew effect based fast method is proposed to generate superposition number matrix of ECG signals.The method is proposed from Matthew effect phenomenon of single beat ECG data trace.In the method,not only the weight of the aggregation points in the matrix is added,its neighborhoods are added too.This improved the construction rate of the stable region.As a result,the stable distribution region of ECG can be expressed by using a small amount of ECG data,and the stable distribution features can be extracted from it.With the same data set in the experiment,similar results were gotten by only 10(6~10 seconds)heartbeats superposition compare to the original hundreds of ECG data.The experimental results show that on the condition of False Positive Rate is zero,the average True Positive Rate reached more than 80% steadily.From this result,ECG signal,as a kind of biometric authentication material,is developing from theoretical research to real scene application,and ECG based authentication becoming a reality into the biometric authentication field.