289 / 1971-01-01 00:00:00

Embedded Face Recognition System Based On Compressive Features

5780

An embedded access control system based on face recognition is presented. The hardware of our system contains an ARM processor (S5PV210) and a normal USB camera. Compared with DSP and FPGA, the S5PV210 professor is with lower cost, good transplant, huge memory and higher computing speed. The face recognition system is implemented by using the Linux operating system, QT graphical interface and Opencv library. To recognize face, compressive sensing features are extracted for representing face by applying an measurement matrix on the original face image. The obtained compressive features maintain the most important information of face image, and they has a powerful discrimination ability. Compared with the original face image, the compressive features only need small memory which can save computing time. Gradient projection method is employed for finding the class of the test image in the face database. The experimental results show that our system is improved in terms of computing time and recognition rate.