Date of Award
Doctor of Philosophy (PhD)
Due to increasing requirements for security, the application and importance of biometrics is growing at a rapid pace. Biometrics is the science of using physiological or behavioral characteristics to determine or verify attributes of a person, including identity. Fingerprints, iris scans, face images, and retina scans are examples of measurements of physiological characteristics that have been proposed and are being used as biometrics. Gait and signature are two primarily behavioral characteristics that have been explored for their use as biometrics. More biometric systems are under development as current biometric technologies satisfy those attributes with mixed success. In biometric recognition, two key properties for useful biometrics are their ability to distinguish among individuals and their stability over time.
A novel approach of measuring carotid pulse signals via a laser Doppler vibrometer remotely is proposed. Laser Doppler Vibrometry (LDV) is used to sense vibration on the surface of the skin above the carotid artery. This motion is related to arterial wall movements associated with the central blood pressure pulse. The non-contact basis of the LDV method has several potential benefits related to non-intrusiveness. To enhance the technical quality of the laser signal during this developmental effort, a small patch (1 cm2) of reflective tape was affixed to the recording site.
The biometric capabilities of Laser Doppler Vibrometry (LDV) signals are evaluated. Several recognition methods are proposed that use the temporal and/or spectral information in the signal to assess biometric performance both on an intra-session basis, and on an inter-session basis involving testing repeated after delays of 1 week to 6 months. A waveform decomposition method that utilizes principal component analysis is used to model the signal in the time domain. Authentication testing for this approach produces an equal-error rate (EER) of 0.5% for intra-session testing. However, performance degrades substantially for inter-session testing, requiring a more robust approach to modeling. Improved performance is obtained using techniques based on time-frequency decomposition, incorporating a method for extracting informative components. Biometric fusion methods including data fusion and information fusion are applied to train models using data from multiple sessions. As currently implemented, this approach yields an inter-session EER of 6.3%.
LDV biometric performance under moderate exercise is tested. A protocol is set up to produce changes in heart rate by physical exercise. Spectrogram based approaches are applied with an EER of 3.6% for inter-state tests, indicating that the LDV pulse signal is stable after moderate physical exercise. The performance degrades during exercise, but improves within 30 seconds as the heart rate recovers during the rest period. The results suggest that the variability caused by heart rate fluctuations and respiration changes decreases within a short time.
Available for download on Sunday, December 15, 2109