TY - GEN
T1 - ECG identification based on PCA-RPROP
AU - Yu, Jinrun
AU - Si, Yujuan
AU - Liu, Xin
AU - Wen, Dunwei
AU - Luo, Tengfei
AU - Lang, Liuqi
N1 - Publisher Copyright:
© Springer International Publishing AG 2017.
PY - 2017
Y1 - 2017
N2 - With the quick development of information technology, people pay more and more attention to information security and property safety, where identity is one of the most important aspects of information security. Compared with the traditional means of identification, biometrics recognition technology offers greater security and convenience. Among which, electrocardiogram (ECG) human identification has been attracted great attention in recent years. As a new type of biometric feature authentication technology, the feature selection and classification of ECG has become a focus of the research community. However, there exist some problems that can impair the efficiency and accuracy of ECG identification, including information redundancy and high dimensionality in feature extraction, and insufficient stability in classification. In order to solve the problems, in this paper, we propose a recognition method based on PCA-RPROP. In this method, firstly, only R points are located to get the original single-cycle waveforms. Then, PCA and whitening are used to process original data, where whitening is to make the input less redundant and PCA is to reduce its dimensionality. Finally, the resilient propagation (RPROP) algorithm is used to optimize the neural network and establish a complete recognition model. In order to evaluate the effectiveness of the algorithm, we compared the PCA feature with the wavelet decomposition and multi-point localization features in an ECG-ID database, and also compared RPROP with traditional BP algorithm, SVM and KNN. The experimental results show that this method can improve the performance compared with other classifiers, and simultaneously reduce the complexity of localization and the redundancy of features. It is superior to the other methods both speed and accuracy in recognition, especially when compared with the traditional BP. It can solve the problems of traditional BP with 2.4% higher recognition accuracy than LIBSVM, and 14 s faster than KNN in terms of time efficiency. Therefore, it is an efficient, simple and practical recognition algorithm.
AB - With the quick development of information technology, people pay more and more attention to information security and property safety, where identity is one of the most important aspects of information security. Compared with the traditional means of identification, biometrics recognition technology offers greater security and convenience. Among which, electrocardiogram (ECG) human identification has been attracted great attention in recent years. As a new type of biometric feature authentication technology, the feature selection and classification of ECG has become a focus of the research community. However, there exist some problems that can impair the efficiency and accuracy of ECG identification, including information redundancy and high dimensionality in feature extraction, and insufficient stability in classification. In order to solve the problems, in this paper, we propose a recognition method based on PCA-RPROP. In this method, firstly, only R points are located to get the original single-cycle waveforms. Then, PCA and whitening are used to process original data, where whitening is to make the input less redundant and PCA is to reduce its dimensionality. Finally, the resilient propagation (RPROP) algorithm is used to optimize the neural network and establish a complete recognition model. In order to evaluate the effectiveness of the algorithm, we compared the PCA feature with the wavelet decomposition and multi-point localization features in an ECG-ID database, and also compared RPROP with traditional BP algorithm, SVM and KNN. The experimental results show that this method can improve the performance compared with other classifiers, and simultaneously reduce the complexity of localization and the redundancy of features. It is superior to the other methods both speed and accuracy in recognition, especially when compared with the traditional BP. It can solve the problems of traditional BP with 2.4% higher recognition accuracy than LIBSVM, and 14 s faster than KNN in terms of time efficiency. Therefore, it is an efficient, simple and practical recognition algorithm.
KW - ECG
KW - Identity recognition
KW - Neural network
KW - PCA dimensionality reduction
KW - RPROP
KW - Whitening
UR - http://www.scopus.com/inward/record.url?scp=85025121628&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-58466-9_37
DO - 10.1007/978-3-319-58466-9_37
M3 - Published Conference contribution
AN - SCOPUS:85025121628
SN - 9783319584652
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 419
EP - 432
BT - Digital Human Modeling
A2 - Duffy, Vincent G.
T2 - 8th International Conference on Digital Human Modeling and Applications in Health, Safety, Ergonomics, and Risk Management, DHM 2017, held as part of 19th International Conference on Human-Computer Interaction, HCI 2017
Y2 - 9 July 2017 through 14 July 2017
ER -