ECG-based Biometric Authentication using Empirical Mode Decomposition and Support Vector Machines
Electrocardiogram (ECG) is an electric signal of cardiac activity posing highly discriminative properties related to human recognition. ECG based authentication has gained much success in recent times however discriminant feature extraction and efficient pattern classification still encounter numerous challenges. This paper proposed a novel methodology for ECG based biometric authentication system. Proposed method first denoise single lead raw ECG signal through empirical mode decomposition (EMD). Region of interest from ECG signals having maximum characteristic information related to subject’s recognition is also extracted through EMD. Next, feature extraction is performed by combination of five features from statistical, time and frequency domains. Finally, selected features were categorized with range of different classification methods such as Support Vector Machines (SVM), K-nearest neighbor (KNN) and Decision Tree (DT). 10-fold cross validation based classification evaluation reveals that SVM with cubic kernel achieves best accuracy of 98.7%, sensitivity of 100% and 98.8% specificity for successful classification of 14 subjects.
Paper
Full text
ECG-based Biometric Authentication using Empirical Mode Decomposition and Support Vector Machines
Semantic Scholar · Computer Science · 2019
Abstract
Electrocardiogram (ECG) is an electric signal of cardiac activity posing highly discriminative properties related to human recognition. ECG based authentication has gained much success in recent times however discriminant feature extraction and efficient pattern classification still encounter numerous challenges. This paper proposed a novel methodology for ECG based biometric authentication system. Proposed method first denoise single lead raw ECG signal through empirical mode decomposition (EMD). Region of interest from ECG signals having maximum characteristic information related to subject’s recognition is also extracted through EMD. Next, feature extraction is performed by combination of five features from statistical, time and frequency domains. Finally, selected features were categorized with range of different classification methods such as Support Vector Machines (SVM), K-nearest neighbor (KNN) and Decision Tree (DT). 10-fold cross validation based classification evaluation reveals that SVM with cubic kernel achieves best accuracy of 98.7%, sensitivity of 100% and 98.8% specificity for successful classification of 14 subjects.