TY - GEN
T1 - Electrocardiographic signal as a biometrie feature
AU - Camacho-Perea, Alejandra
AU - Maya-Martinez, Sergio U.
AU - Tovar-Corona, Blanca
AU - De La Rosa-Montero, Erick D.
N1 - Publisher Copyright:
© 2017 IEEE.
PY - 2017/11/14
Y1 - 2017/11/14
N2 - This work proposes a biometrie identification system based on electrocardiographic signal, for online applications. To acquire the signal, the NI MyDAQ and the AD8232 Heart Rate Monitor module were used. Once the signal was digitized, filtering was applied and algorithms to process the signal were designed to detect the fiducial markers of the signal, such as the R peak, the Q wave, the ending of the S wave and the ending of the T wave, in order to segment the signal, so it could be normalized on amplitude and time. This processing was tested over a data base with 3701 samples of ECG signals from 20 volunteers, generated for this work. The signals were taken using the standard bipolar lead I. Two perceptron neural networks were developed to classify the samples. The first one uses as a characteristic pattern, the PQRST cycle while the second one uses the first 70 coefficients of the Discrete Cosine Transform (DCT) of the characteristic PQRST cycle. The performance of both classifiers was evaluated in different scenarios: off line without intruder, off-line with intruder, on-line without intruder. On the offline tests the classification based on the complete cardiac cycle had an accuracy of 99.29%. The classification based on the first 70 coefficients of the Discrete Cosine Transform, had an accuracy of 97.89%. The Online tests had an accuracy of 98.33% and 90.83 % respectively.
AB - This work proposes a biometrie identification system based on electrocardiographic signal, for online applications. To acquire the signal, the NI MyDAQ and the AD8232 Heart Rate Monitor module were used. Once the signal was digitized, filtering was applied and algorithms to process the signal were designed to detect the fiducial markers of the signal, such as the R peak, the Q wave, the ending of the S wave and the ending of the T wave, in order to segment the signal, so it could be normalized on amplitude and time. This processing was tested over a data base with 3701 samples of ECG signals from 20 volunteers, generated for this work. The signals were taken using the standard bipolar lead I. Two perceptron neural networks were developed to classify the samples. The first one uses as a characteristic pattern, the PQRST cycle while the second one uses the first 70 coefficients of the Discrete Cosine Transform (DCT) of the characteristic PQRST cycle. The performance of both classifiers was evaluated in different scenarios: off line without intruder, off-line with intruder, on-line without intruder. On the offline tests the classification based on the complete cardiac cycle had an accuracy of 99.29%. The classification based on the first 70 coefficients of the Discrete Cosine Transform, had an accuracy of 97.89%. The Online tests had an accuracy of 98.33% and 90.83 % respectively.
KW - Biometrics
KW - DCT
KW - ECG
KW - Electrocardiogram
KW - Neural Network
KW - Perceptron
UR - http://www.scopus.com/inward/record.url?scp=85040545111&partnerID=8YFLogxK
U2 - 10.1109/ICEEE.2017.8108877
DO - 10.1109/ICEEE.2017.8108877
M3 - Contribución a la conferencia
AN - SCOPUS:85040545111
T3 - 2017 14th International Conference on Electrical Engineering, Computing Science and Automatic Control, CCE 2017
BT - 2017 14th International Conference on Electrical Engineering, Computing Science and Automatic Control, CCE 2017
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 14th International Conference on Electrical Engineering, Computing Science and Automatic Control, CCE 2017
Y2 - 20 September 2017 through 22 September 2017
ER -