Electrophysiological Multimodal Biometrics

We study the potential of human identification using electrocardiography (ECG) and electrodermal activity (EDA) data, collected during a set of activities designed to stimulate different emotional responses in the subjects. The methodologies for the integration of biosignals, extracted features and the classification combination will be presented as the base setting for enabling multimodal continuous biometric systems. In this contribution we describe the collected biosignals, feature selection, and pattern recognition techniques, associated with the applicability of these signals in a multimodal biometrics framework. Experimental evaluation and results is also presented using real-­‐world electrophysiological signals. Our results compare favorably to standard behavioral biometric techniques, where within a one minute acquisition we obtain 1% equal error rate classification performance. The current research opens the possibility of continuous multimodal biometric systems that can combine information from hard and soft biometric traits.