ECG Identification System Using Neural Network with Global and Local Features

This paper proposes a human identification system via extracted electrocardiogram (ECG) signals. Two hierarchical classification structures based on global shape feature and local statistical feature is used to extract ECG signals. Global shape feature represents the outline information of ECG signals and local statistical feature extracts the information between signals in time domain. Genetic algorithm based back propagation neural network is used as the specific classifier. Experiment results show that our identification system can achieves an average 97.6% accuracy on a 38 subjects of PTB public ECG database and an average 100% accuracy on an 18 subjects of MIT-BIH public ECG database, which demonstrates the proposed system can reach satisfactory effects. [For full proceedings, see ED571459.]