ECG based User Authentication for Wearable Devices using Short Time Fourier Transform

Abstract

Electrocardiogram (ECG) is a promising biometric. There has been much research on ECG based user authentication and identification, but there have been few works to investigate ECG biometrics for stand-alone wearable ECG sensors, for quick response time using a single pulse ECG, and for small wearable devices that may have limited access to others’ ECG information. Recently, ECG user authentication method using spectrogram yielded excellent detection performance. However, spectrogram only utilizes magnitude of short time Fourier transform (STFT) and phase information was overlooked for ECG features. In this paper, we address the issues of wearable ECG sensors, quick response time, and limited access to others’ ECG information using a new STFT based method that uses phase information. Our proposed method yielded 0.9% EER for ECG data set from wearable ECG sensors (15 subjects) and 2.2% EER (equal error rate) for public ECG-ID database (89 subjects).