ECG-based biometrics using convolutional neural networks and ensemble empirical mode decomposition
Conference: BIBE 2019 - The Third International Conference on Biological Information and Biomedical Engineering
06/20/2019 - 06/22/2019 at Hangzhou, China
Proceedings: BIBE 2019
Pages: 5Language: englishTyp: PDF
Personal VDE Members are entitled to a 10% discount on this title
Authors:
Huang, Runxin; Deng, Yanjun (College of Electronic and Communication Engineering, Hangzhou Dianzi University, Hangzhou, China)
Zhao, Zhidong (College of Electronic and Communication Engineering, Hangzhou Dianzi University, Hangzhou, China & Hangdian Smart city research center of Zhejiang Province, Hangzhou Dianzi University, Hangzhou, China)
Guo, Chunwei (Hangdian Smart City Research Center of Zhejiang Province, Hangzhou Dianzi University, Hangzhou, China)
Abstract:
Electrocardiogram (ECG) biometric technology is being widely studied for probable use in identification and authentication tasks. To solve the these two problems , we propose novel two-stage ECG biometric methods that are based on Ensemble Empirical Mode Decomposition (EEMD) and Convolutional Neural Networks (CNN). A blind segmentation technique effectively avoids the localization defect for ECG fiducial characteristic points. The random ECG segmentations are decomposed by EEMD to obtain 2D Time-Frequency (TF) representation, which is used as the input to train the CNN model. CNN can adaptively capture the representative features for the identification and authentication task, which is free of human interference with no manual feature extraction. Support Vector Machine(SVM) and cosine similarity are used for classification identification and authentication matching of feature vectors, respectively. The proposed algorithm performances are tested on the European ST-T database (EDB). The Equal Error Rate (EER) and the Identificaion Rate(IR) of EDB are 3.85%, 99.1%, respectively. The result indicates the feasibility and practicability of our proposed methods for biometric tasks.