A novel framework for ECG biometric verification on mobile devices utilising activity classification

Considering the digital world, almost all online transactions (e.g. payments, shopping, etc.) need identity verification and information security. In addition to models such as traditional passwords and PINs, biometrics such as facial recognition and fingerprints have also begun to be used frequently for this purpose. The literature has often stated that electrocardiogram (ECG) biometrics can also provide reliable results and increase performance in multi-models. However, it has also been stated that the characteristics of ECG signals change over time due to environmental, biological or physiological reasons such as physical activities and emotional states. This affects the performance and stability of the model.

Another open challenge is the need for difficult-to-use devices and sensors to collect ECG data. With the development of wearable devices, many studies have evaluated the performance of these devices. However, to provide a reliable suitable real-life scenarios ECG-based biometric verification model, many parameters must be investigated in depth.

The scope of this thesis is to examine the parameters affecting ECG biometrics in verification models, to create a novel framework to increase long-term stability and to test the created framework on various mobile devices.

The contribution of this thesis to the literature is by creating an activity-aware and emotional status-aware biometric verification framework, demonstrating the usability of this framework for medical and wearable devices.