Low resolution and un-sharp facial images are always captured from surveillance videos because of long human-camera
distance and human movements. Previous works addressed this problem by using an active camera to capture close-up
facial images without considering human movements and mechanical delays of the active camera. In this paper, we
proposed a unified framework to capture facial images in video surveillance systems by using one static and active
camera in a cooperative manner. Human faces are first located by a skin-color based real-time face detection algorithm.
A stereo camera model is also employed to approximate human face location and his/her velocity with respect to the
active camera. Given the mechanical delays of the active camera, the position of a target face with a given delay can be
estimated using a Human-Camera Synchronization Model. By controlling the active camera with corresponding amount
of pan, tilt, and zoom, a clear close-up facial image of a moving human can be captured then. We built the proposed
system in an 8.4-meter indoor corridor. Results show that the proposed stereo camera configuration can locate faces with
average error of 3%. In addition, it is capable of capturing facial images of a walking human clearly in first instance in
90% of the test cases.
This paper describes an embedded multi-user login system based on fingerprint recognition. The system, built using the Sitsang development board and embedded Linux, implements all fingerprint acquisition, preprocessing, minutia extraction, match, identification, user registration, and template encryption on the board. By careful analysis of the accuracy requirement as well as the arithmetic precision to be used, we optimized the algorithms so that the whole system can work in real-time in the embedded environment based on Intel(R) PXA255 processor. The fingerprint verification, which is the core part of the system, is fully tested on a fingerprint database consists of 1149 fingerprint images. The result shows that we can achieve an accuracy of more than 95%. Field testing of 20 registered users has further proved the reliability of our system. The core part of our system, then embedded fingerprint authentication, can also be applied in many different embedded applications concerning security problems.
This paper presents an experimental study of the implementation of a face authentication system for mobile devices. Our system is based on a widely adopted face recognition technique called Principal Component Analysis (PCA). The execution time of the baseline system on a PDA is unacceptably slow -- a typical authentication session takes more than 30 seconds. To make real-time face authentication possible on mobile devices, optimization is needed. In our study, extensive profiling is done to pinpoint the execution hotspots in the system. Based on the profiling results, our optimization strategy focused on replacing the large amount of slow floating point calculations with their fixed-point versions. Range estimation is also carried out to determine the range of floating point values that must be accommodated by the final, fixed-point version of our system. Compared with the baseline system, experimental results indicate that our optimized system runs as much as 47 times faster for PCA projection. Using the optimized system, a complete authentication session takes only 5 seconds. Real time face authentication for mobile device is achieved with no significant loss in recognition accuracy.