Biomedical Signal Processing and Control 76 (2022) 103692 Available online 13 April 2022 1746-8094/© 2022 Elsevier Ltd. All rights reserved. A wavelet-based capsule neural network for ECG biometric identifcation Imane El Boujnouni a, * , Hassan Zili b , Abdelhak Tali a , Tarik Tali c , Yassin Laaziz a a Laboratory of Information and Communication Technologies, Abdelmalek Essaadi University, Tangier, Morocco b Laboratory of Computer Science, Systems and Telecommunications, Abdelmalek Essaadi University, Tangier, Morocco c CEO at Taliware.com, California, USA A R T I C L E INFO Keywords: Electrocardiogram Biometric identifcation Wavelet transform Capsule network ABSTRACT Electrocardiogram (ECG) signals have received a high level of attention from the biometric research community due to their unique nature for each person, which makes them suitable for developing accurate and reliable human identifcation systems. Although most existing ECG-based biometric recognition methods have received prominent results, several consecutive heartbeat segments are used in their approaches to achieve high accuracy, which is challenging to apply in biometric systems deployed in real-world applications. This paper proposes a new approach for human identifcation via ECG, based on a combination of Continuous Wavelet Transform (CWT), Discrete Wavelet Transform (DWT) along with a novel kind of deep learning technique known as Capsule network. The CWT is used to transform a single heartbeat signal into the time–frequency domain, and the DWT is adopted to extract spectral information of 2D frequency-time scalogram images to further improve accuracy. The discrete wavelet coeffcients are then used as input to the capsule network promoting the recognition perfor- mance due to its high learning capacities. To support real-life practicality of ECG biometric identifcation system, the effectiveness and effciency of our approach were evaluated over four databases that include normal and abnormal ECG records: PTB Diagnosis ECG (PTB), MIT-BIH Arrhythmia, MIT-BIH Normal Sinus Rhythm (NSRDB), and the MIT-BIH ST Change (STDB) databases. Experimental results demonstrate that our proposed method was able to achieve high identifcation accuracies and outperforming other state-of-the-art methods, by achieving an accuracy of 99.5%, 98.1%, 98.2%, and 100% on the PTB, MIT-BIH arrhythmia, STDB, and NSRDB respectively. Furthermore, the approach showed very good generalization ability since the training and test sets were completely different, which demonstrates the feasibility to promote the application of our approach in practice. 1. Introduction Biometric identifcation refers to the automated method to identify a person based on physiological and/or behavioral human characteristics. The most well-known forms of biometric modalities that have been successfully deployed in practical applications include the face, fnger- print, speech, handwriting, and iris. However, extensive research has shown that they are more likely to be hacked and manipulated by impersonation attacks [1]. In recent years, physiological signals such as electroencephalography (EEG) [2,3] electromyography (EMG) [4,5], electrocardiogram (ECG) [6–8], and photoplethysmography (PPG) [9,10], have demonstrated strong potential in biometric recognition systems. Among these physiological signals, the ECG signal, which is a recording of electrical activities of the heart in a non-invasive way, has been demonstrated to have the needed characteristics such as unique- ness, universality, liveness detection, permanent acquisition for a robust and reliable recognition process. Furthermore, its acquisition becomes convenient as it can be easily obtained from the fngers of a user [11]. Notwithstanding, ECG biometrics are faced with challenges affecting the morphology features of ECG waveforms, such as noise, movement arti- facts, intra-subject and inter-subject variability [12,13]. The inter-subject variability (Uniqueness) which is referred to the variability between cardiac signals from different subjects, appears due to the geometrical and physiological differences of heartbeat including heart size, the orientation of heart muscle, the electric conductivity, and order of activation of cardiac muscle [14]. The intra-subject variability (Permanence) which represents the variability between different cardiac signals from the same person, ap- pears due to several sources including the heart rate variability, chest * Corresponding author. E-mail address: imane.elboujnouni@etu.uae.ac.ma (I. El Boujnouni). Contents lists available at ScienceDirect Biomedical Signal Processing and Control journal homepage: www.elsevier.com/locate/bspc https://doi.org/10.1016/j.bspc.2022.103692 Received 11 November 2021; Received in revised form 11 March 2022; Accepted 9 April 2022