Effective Video Processing Architectures for Face Recognition Algorithms Implementation

The paper presents a study of autonomous face recognition systems based on high performance DSP, so called Media Processors and on Field Programmable Gate Array (FPGA) devices. An overview of the most powerful Media Processors and FPGAs available today is followed by an analysis of the common features of the face recognition algorithms used to implement an application. The paper presents also a generic system implemented using DSP Media processors, both from a hardware or software perspective. The presentation is focused on the possibility of implementation for face recognition algorithms in embedded systems, taking as examples databases with a limited number of faces. Results, estimations and benchmarks for the generic systems are also presented at the end of the paper.