Electrocardiogram for Biometrics by using Adaptive Multilayer Generalized Learning Vector Quantization (AMGLVQ): Integrating Feature Extraction and Classification

摘要

Electrocardiogram (ECG) signal for human identity recognition is a new area onbiometrics research. The ECG is a vital signal of human body, unique, robustness to attack,universality and permanence, difference to others traditional biometrics technic. This study alsoproposes Adaptive Multilayer Generalized Learning Vector Quantization (AMGLVQ), thatintegrating feature extraction and classification method. The experiments shown that AMGLVQcan improve the accuracy of classification better than SVM or back-propagation NN and also ableto handle some problems of heartbeat classification: imbalanced data set, inconsistency betweenfeature extraction and classification and detecting unknown data on testing phase.