Research Of Personal Identification Algorithm Based On The Analysis Of Finger ECG Signal

In a highly informative modern society,with the rapid development of transportation,network and communication,the scope of human activities is growing.There are more and more serious security problems when involving individual identification.Biometrics is a kind of identity recognition technology via utilizing some unique and long-term stability biological characteristics,such as iris,fingerprint,face,voice,handwriting,gait,etc.They all have high accuracy and reliability identification.However,there are still some security issues occurring in biometrics,like fake fingerprint,false iris,imitation of handwriting and so on.So several novel biometric identification technologies with high security is urgently required.Electrocardiogram(ECG)contains a unique identification information of the human body,and has outstanding feature of high security.Accompanying with the emergence of ECG chips with the features of small size,lower power and easy integration recently,acquisition of the ECG signal from fingers can be realized,which can be widely applied to family life and become more convenient and efficient to achieve individual identification.The paper presented the research of personal identification algorithm based on the analysis of finger ECG signal.First,denoised the finger ECG signal by wavelet soft threshold and genetic algorithm.Second,the sparse characteristics are studied,the finger ECG identification algorithm both based on the sparse coding under KSVD+PCA and based on the improved label consistent LC-KSVD are put forward.Finally,the proposed finger ECG identification algorithms are implemented in two finger ECG database and achieved recognition rate.The main ideas of this paper are mainly as follows:1.This paper describes the generation mechanism of the ECG signals,waveform characteristics,ECG acquisition mode and the development of identity recognition algorithm based on Finger-ECG.Introduced the evaluation index of ECG recognition and two finger-ECG databases,which provide a theoretical basis for finger ECG signal when applied to the field of individual identification.2.The finger ECG identification algorithm based on the sparse coding under KSVD+PCA is put forward.First,the finger ECG is denoised by wavelet soft threshold and genetic algorithm.The R-peak detection,single cycle division and normalization are performed to get the single-period wave group of P-QRS-T.Combine with finger ECG characteristics,P-QRS-T wave groups were extracted as constitute characteristic to build the dictionary model and then be trained to be redundant dictionary by KSVD+PCA.Then every part of the feature vecors is used for sparse coding,implement the sparse representation in the dictionary.The last,test the algorithm performance utilizing two finger ECG database(CYBHi,Surface ECG data)and obtain the recognition rate of 98.333% and 100%.3.A new finger ECG identification algorithm based on improved label consistent LC-KSVD is put forward.First,P-QRS-T wave groups were extracted as training sample.Then put forward the adaptive child dictionary and adjustable labels to improve the consistency of the LC-KSVD(Label consistent KSVD,LC-KSVD).Second,use the improved LC-KSVD1 and LC-KSVD2 complete the algorithm recognition.In the objective function,the K-SVD algorithm is used to combine the discriminant error,the reconstruction error and the classification error together to update the dictionary and train a classifier while the label information between the input signal and the dictionary atoms is one-to-one correspondence with each other.Finally,test the algorithm performance utilizing two finger ECG databases(CYBHi,Surface ECG data)and obtain the recognition rate of 99% and 100%.In this paper,algorithms for personal identification based on finger ECG analysis are proposed,which lays a theoretical foundation and technical support for the practical application of finger ECG identification technology.