Verification of individuals from accelerometer measures of cardiac chest movements

Abstract

Biometric verification is gaining popularity particularly for personal security during internet and mobile device usage. A novel approach for verification of individuals is proposed to measure mechanical cardiovascular activity through an accelerometer sensor placed on the surface of the chest above the sternum. Time frequency analysis methods are employed to evaluate biometric performance. Accelerometer measurements were acquired on two different sessions from ten subjects after delays ranging from 1 to 2 weeks. For individual subject verification, Gaussian mixture models were built per each individual and a background model was created for the remaining impostors. A likelihood ratio test with background model was employed for testing. In this study we found preliminary evidence for the use of the cardiovascular signal measured with an accelerometer placed on the sternum as a biometric sensor to verify individuals. Verification testing using this approach obtained a mean EER rate of 0.06 for inter-session testing.