Low Cost Fpga-Based Highly Accurate Face Recognition System using Combined Wavelets with Subspace Methods

Abstract

Low cost implementation of an accurate face recognition system is useful in many applications. A face recognition algorithm based on Daubechies wavelets is proposed. The algorithm uses each of ICA or PCA for dimensionality reduction. Wavelets are employed to increase the discriminative power of subspace methods. Effectiveness of using wavelets is shown by performing comparison between proposed algorithm and other solutions. Complete hardware implementation of the algorithm is then proposed. It is shown that the entire face recognition system can be implemented on a low cost 300 K gates Spartan-II FPGA. The system works with high frame per second processing rate and only half of the logic resources on FPGA are consumed by face recognition logic.