This chapter discusses the feasibility of using the Electrocardiogram (ECG) for human identification. ECG falls under the umbrella of medical biometrics, i.e., physiological signals that are typically used for disease diagnosis, but also carry subject discriminative information. As opposed to traditional static biometric modalities like the iris, the fingerprint or the face, ECG is a time dependent signal affected both by physical and emotional activities. Therefore, one of the challenges that are studied in this work is the design of permanent and, at the same time discriminative features, which are robust to heart rate changes.

A feature extraction methodology based on the autocorrelation of ECG recordings is presented and evaluated on a public database. In addition, the advantages of using the standard 12 lead ECG system in the recognition process are discussed, and various fusion strategies are presented. Experimental results indicate increased recognition accuracy for the fused case (100% for 14 subjects).

Furthermore, this chapter advocates ECG biometrics as a natural choice for handling medical information. Given the medical origin of the signal and the fact that in a clinical setting this measurement is collected irrespective of the recognition task, an ECG based biometric signature can be used to manage the patient's information. A body area network (BAN) is described as the application environment, while the findings can be generalized for a wide range of clinical settings.