Capsule Network for the Identification of Individuals Using Quantized ECG Signal Images

Electrocardiogram (ECG)-based biometric systems are popular due to their uniqueness and tamper-resistance. The author has designed a digital signal-based deep learning technique for person identification using the ECG signal, in this letter. The temporal variations are very high in the traditional ECG signal. Hence, processing and computational costs are very high for this nonstationary signal. The author converted the analogue signal into a quantiszed signal representation to minimize the temporal variations. These quantized representations are transformed into 128×128 images and utilized as input to the capsule network for further identification. The proposed methodology improves the identification of the beat variations from person to person. As a result, the algorithm was able to operate at a lower cost and with greater speed. The proposed model does not require any manual extraction of fiducial points. The performance of the proposed model is tested on the ECG-ID, Physikalisch-Technische Bundesanstalt, Check Your Bio-signals Here Initiative, and University of Toronto databases. The proposed system can identify a subject within 0.3365 s. The experimental results show that the proposed end-to-end identification system performs better than the earlier state-of-the-art techniques.