다선형 부분공간 학습방법을 이용한 ECG기반 생체인식

Abstract

Biometric techniques using biometric information such as fingerprints, faces, and iris are widely used in everyday life. Biometrics using physical features are dangerous because of counterfeiting and tampering. Biological signals using electrical bio-signals are highly secure. In recent years, we are actively studying bio-signals using electrical bio-signals. In this paper, we analyze the performance of biometrics using multilinear subspace learning from electrocardiogram information. We compare the performance using MPCA, MLDA in order to demonstrate superior performance over the existing subspace learning techniques such as principal component analysis and linear discriminant analysis. We confirmed that 97% recognition rate is obtained when applying the multilinear subspace learning by applying Physionet's MIT-BIH database.