Aziz Umit Batur and Bruce E. Flinchbaugh
Texas Instruments

Face recognition is an important part of today's emerging biometrics and video surveillance markets. As face recognition algorithms move from research labs to real world products, power consumption and cost become critical issues, and DSP-based implementations become attractive. Our goal in this research project was to evaluate the CPU and memory requirements of face recognition algorithms on the TMS320C64x platform to determine the feasibility of implementing DSP-based face recognition systems. The results of our project demonstrate that a generic C implementation with a modest C level optimization effort results in a face recognition software prototype that has low CPU and memory requirements on TMS320C64x, and runs with high speed to enable real-time applications. Therefore, well-optimized face recognition implementations on TMS320C64x are an effective design choice for embedded face recognition systems.