Face Biometric Applications

For general background information see Overview of Biometric Projects.

This project involves the use of one or more face biometric systems in several applications. While the techniques are being implemented in the companion project, your tasks will be to

PDA User Authentication - Robert Zack

Mobile computing devices are increasingly used to access security sensitive applications and data. This project is intended to build upon the existing body of biometric knowledge to compare facial recognition techniques to other biometric systems and determine if facial recognition is feasible using the capabilities of personal digital assistants (PDA) and laptop/notebook computers. The objectives of this project are to determine if facial recognition is feasible and secure when used to authenticate users of personal digital assistants.

While facial recognition has been studied in-depth, additional work is needed to explore the viability of using facial biometrics to secure access to PDA’s and security sensitive applications that run on them. Limited processing capabilities, networking, storage, video capture, and other constraints make this a challenging problem to solve.

The primary objective of this project is to develop an application to secure access to a mobile computing device and to conduct experiments that measure performance and accuracy of the security solution. A preliminary analysis of the problem and the literature search should, for example, determine whether the face biometric is the most appropriate biometric for this application. The data collected from the experiments will help to develop a technology acceptance model (TAM) to determine if this security solution is acceptable to secure PDA devices.

Longitudinal Face Biometric (Authentication or Identification) Study - Fred Penna

Most biometric studies use samples, such as face photos, taken over a short period of time, and these samples are used for both enrollment (training) and testing (authentication or identification). However, a person's face changes over time as he/she ages. Correspondingly, the performance (accuracy) of a biometric system degrades over time, and this degradation has been little studied. Such studies would, for example, train the system on old samples (such as face images taken 10 years ago) and test the system on new ones (taken more recently).

Your first task is to gather a database of photos of, say, ten people over a time period of at least 10 years, preferably 20-40 years. For example, you might gather from older relatives' photos that were taken over a 20-40 year time span. To allow for training then recognition experiments, ideally you should have 5-10 photos of an individual at every 10-year interval.

An interesting experiment would be to obtain recognition performance (accuracy) at the same time (+/- 1 year), at a 10-year interval, at a 20-year interval, at a 30-year interval, etc.