Effect of Body Posture on Non-Fiducial Electrocardiogram Based Biometric System

The electrocardiogram (ECG) can serve as a useful biometric; however, the ECG may be susceptible to morphological changes due to changes in body posture. The objective of this paper is to study the effect of body posture on a non-fiducial biometric system. ECG data were collected from 52 subjects over multiple sessions in the supine, sit, and stand postures; each posture was held for approximately 60 seconds. The postures' intervals were divided into 20 s segments with 10 s overlap which generated 417 segments. Three biometric system variants were considered to study the effect of posture: 1) single-posture system, 2) multi-posture system, and 3) posture-aware system. The singleposture system variant only compares test segments to single posture templates in the biometric database. The multi-posture system variant has templates from supine, sit, and stand postures. The posture-aware system variant uses the subject's posture and then compares incoming test segments with the templates from the matching posture. The single-posture system variant had the highest average true positive (TP) rate of 65.5% in the supine posture and the lowest average TP rate of 50.8% in the stand posture. The multi-posture system and posture-aware system variants had an average TP rate of 81.5% and 80.8%, respectively. The multi-posture system variant identified the posture of the test segments with an average rate of 71.0%. Results indicated that posture affects the ECG measurement; therefore, ECG biometric systems should take this into consideration in their design and implementation.