A novel personal identification system using doorknob lead electrocardiograms for unconscious authentication in...

In highly information-oriented society, personal authentication technology is essential. Biometric authentication is becoming popular as a method of personal authentication from the viewpoint of usability. In this research, in order to realize unconscious personal authentication during daily activities, we proposed a novel biometric authentication system using a doorknob-type electrocardiogram (ECG) measuring device. In our previous study, it was shown that ECG obtained with a contact-type electrode on doorknob and a capacitive-type electrode on the floor could be used for personal identification. However, identification performance is easily affected by noise from body movements and other factors, due to loose contact between electrodes and the body.

In this paper, we proposed to add two preprocessing techniques to the system. Synchronized averaging process was applied to the measured ECG waveforms. Then, data augmentation was applied to the machine learning training data.

It was found that synchronized averaging with 5 consecutive wave segment improved accuracy by 10%. It was also found that training data augmentation improved the performance even under limited amount of ECG data.

The results demonstrate that remarkable performance improvement can be achieved even with short term door-knob ECG by using synchronized averaging and data augmentation.