eigenPulse: Robust human identification from cardiovascular function

摘要

This paper presents eigenPulse, a new method for human identification from cardiovascular function. Traditional biometric techniques, e.g. face and fingerprint, have used eigen analysis to exploit databases with tens of thousands of entries. One drawback of traditional biometrics is that the credentials, for example, fingerprints, can be forged making the systems less secure. Previous research [S.A. Israel, J.M. Irvine, A. Cheng, M.D. Wiederhold, B.K. Wiederhold, ECG to identify individuals, Pattern Recognition 38(1) (2005) 138–142] demonstrated the viability of using cardiovascular function for human identification. By nature, cardiovascular function is a measure of liveness and less susceptible to forgery. However, the classification techniques presented in earlier work performed poorly over non-standard electrocardiogram (ECG) traces, raising questions about the percentage of the population that can be enrolled. This paper combines the traditional biometrics’ use of eigen analysis and previous analysis of cardiovascular function to yield a more robust approach. The eigenPulse processing had a near 100% enrollment rate, with a corresponding higher overall performance.