Implementation of a Biometric Identification System using Electrocardiogram ECG/EKG

Abstract

Nowadays, Biometrics is extensively being used for the purpose of authentication in security related aspects. Biometrics deals with individuals’ identification through their physiological characteristics such as fingerprint, Deoxyribonucleic Acid DNA, Electrocardiogram ECG, Face, Voice…etc. Many of these models have limitations and contains: difficultly of extraction (DNA), unique utilization of the hardware (IRIS)…etc. Hence, ECG is chosen for its accuracy, hardware utilization in tele-monitoring, and high security level. In this presented treatise, an ECG based biometric system has been developed. The project is divided into two main parts: Hardware and Software. The hardware part, performs the signal acquisition, amplification, and digitalization. The hardware is provided with the alternative of wired via UART or wireless transmission via XBee RF modules. In software part, the biometric algorithm uses the concepts of machine learning and pattern recognition. Basically, the algorithm is divided into two main part: training (enrollment) phase and testing phase. In training phase, the system gets indoctrinated with a set of ECG data recordings of different people. Then, it extracts their features after pre-processing in a form of spectral information. Features are taken into higher dimensionality space of 28 (256D) and scattered separately to form labeled classes (Principal Components Analysis – PCA). Once the testing data arrives, classification process affects them into their corresponding class, which is identification decision, using the Euclidean distance. The algorithm was developed based on the MIT-BIH Arrhythmia Database and then tested on our customized database acquired using the developed hardware. The two main aspects that have been under focus are the execution time and accuracy or identification rate. An accuracy of 98% has been achieved with the developed system.