ECG Based Biometric using Wavelet Packet Decomposition
Sugondo Hadiyoso1, Achmad Rizal2, Inung Wijayanto3

1Sugondo Hadiyoso, school of applied science, Telkom University, Bandung, Indonesia.
2Achmad Rizal, school of electrical engineering, Telkom University, Bandung, Indonesia.
3Inung Wijayanto, school of electrical engineering, Telkom University, Bandung, Indonesia.
Manuscript received on September 22, 2019. | Revised Manuscript received on October 20, 2019. | Manuscript published on October 30, 2019. | PP: 2178-2183 | Volume-9 Issue-1, October 2019 | Retrieval Number: A9699109119/2019©BEIESP | DOI: 10.35940/ijeat.A9699.109119
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (

Abstract: Biometric technology has been commonly used for authentication. Fingerprint or iris become one of the biometrics that is widely applied. However, this type of biometrics tends to be easily falsified and damaged. So it is misused for manipulating actions and even crime. Therefore a new biometric method is needed to overcome this problem. One potential modality is biometrics based on an electrocardiogram (ECG) signal. This research simulates a one-lead ECG waveform for person authentication. ECG waves were taken from eleven healthy adult volunteers with a length of 60 seconds. ECG waves from each person are segmented into 10 sections so that a total of 110 ECG waves are used for person authentication simulations. All noise of the ECG waves was removed using a bandpass filter to reduce artifacts and high-frequency noise. Wavelet packet decomposition (3 Level) was applied to decompose the signal in several intrinsic parts so that typical wave information can be retrieved. Entropy-based feature extraction applied to all decomposed signals. A total of 14 entropy features have been calculated and used as predictors in the classification process. Validation and performance tests are carried out by cross-validation combined with linear discriminant analysis and support vector machines with five scenarios. The proposed method provides the highest accuracy of 71.8% using discriminant analysis and cubic support vector machine. The best accuracy value was achieved if all entropy features from all wavelet decomposition levels are used as predictors in the classification process. This research is expected to be a reference that ECG has the potential to become a future biometric modality.
Keywords: ECG, person authentication, wavelet, decomposition.