High-Performance Real-Time Face-Detection Architecture for HCI Applications

This paper proposes a novel hardware structure and FPGA implementation method for real-time detection of multiple human faces with robustness against illumination variations and Rotated faces. These are designed to greatly improve face detection in various environments, using the Adaboost learning algorithm and MCT techniques, Rotation Transformation, which is robust against variable illumination and rotated faces. The overall structure of proposed hardware is composed of a Color Space Converter, Noise Filter, Memory Controller Interface, Rotation Transformation, MCT (Modified Census Transform), Candidate Detector/Confidence Mapper, Position Resizer, Data Grouper, Overlay Processor and Color Overlay Processor. The experiment was conducted in various environments using a QVGA Camera, LCD Display and Virtext5 XC5VLX330 FF1760 FPGA, made by Xilinx. Implementation and verification results showed that it is possible to detect at least 32 faces in a wide variety of sizes at a maximum speed of 149 frames per second in real time.