Fast Face Detection in Field Programmable Gate Array

Face detection has been widely used in many fields such as security, intelligent human interface etc., and concerns on how to meet the criterion of miniature and real-time have gradually become independent research topics. Traditionally, systems of face detection/recognition are either implemented in very specific hardware or software system. This paper tailors an algorithm of face detection using skin color and presents an efficient implementation on a single Field Programmable Gate Array (FPGA) chip. The implementation exploits the inherit parallelism of hardware and takes advantages of hardware multiply-accumulate units and shift register (RAM-Based) units on FPGA. The entire system uses a single FPGA coupled to a pair of SDRAM chips, and no extra parts. The system was implemented and tested with face detection on a 1280×1024 frame taking 16.66 ms (60 frames per second ). This system can be used as a reference for lightweight embedded vision systems.