Outlier detection in non-intrusive ECG biometric system

Abstract

A recent trend in the field of biometrics is the use of Electrocardiographic (ECG) signals. One of the main challenges of this new paradigm is the development of non-intrusive and highly usable setups. Fingers and hand palms for example, allow the ECG acquisition at much more convenient locations than the chest, commonly used for clinical scenarios. These new locations lead to an ECG signal with lower signal to noise ratio, and more prone to noise artifacts. In this paper, we propose an outlier removal system to eliminate noisy segments, and enhance the performance of non-intrusive ECG biometric systems. Preliminary results show that this system leads to an improvement on the recognition rates, helping to further validate the potential of ECG signals as complementary biometric modality. © 2013 Springer-Verlag.