Face Recognition Algorithm Based On FPGA And Its Implementation

With the development of Internet,requirements of the information security and personal authentication continue to increase,in this context,the biological recognition concept of our country becomes hot.The fingerprint recognition technology has been developed in the past,face recognition technology with higher recognition rate ushered in the development of good machine.Since 2015,face recognition technology is warming in the science and technology circle continuous,all kinds of face recognition applications appear layer upon layer without end.At present,there are many kinds of algorithms for face recognition,and the theoretical basis of this paper is Gabor wavelet transform and HMM face recognition method.The results show that the proposed algorithm is very high on the FPGA hardware.With the application of digital technology in recent years,technology of FPGA developed fast,it has the characteristics of short development cycle,easy development,users can quickly logical unit within the FPGA to develop their own modules.Therefore,the algorithm of face recognition is realized by FPGA hardware.At the same time,the flexibility of its development also reduces the cost of system design.FPGA of series Cyclone II produced by Cyclone company,especially rich in resources,for the design and implementation of the entire face recognition system provides a guarantee.The specific content of the paper is:(1)extraction algorithm analysis: classical statistical features in the research and implementation of existing algorithms based on the Gabor face recognition algorithm based on wavelet transform and HMM are fused to form a new effective algorithm of face recognition.At the same time,the face recognition technology of support vector machine is studied,and the Gabor kernel function can be used to adjust the kernel function parameters to adapt the kernel function to the features of the face.Furthermore,there is another innovation that others Viterbi segmentation and Baum-Welch parameter estimation in Markov chain requires iterative solution,and the number of evaluation value setting convergence condition,reduces the amount of computation time,and improves the computational accuracy.In this paper,the algorithm is based on Hidden Markov model and Gabor wavelet transform.This algorithm works as follows: firstly,a group of nodes set on the image using the expansion characteristics of the use of Gabor wavelet in different directions and scales;then,all nodes implement dimensionality reduction related to the operation,use of this process is the main element analysis method;the next generation,Gabor face.Finally,in order to get the optimized model parameters in the process of recognition,the observed vector is needed to train the hidden Markov model,and these vectors are all the characteristics of Gabor face.In the course of the study,a total of two hundred images were selected for training,and the rest of the two hundred images were used for identification.Divided into two groups of experiments,the first group of experiments is to treat the recognition of image recognition;the second experiment is the first to treat the recognition image by hand to carry out partial occlusion,and then recognition.The experimental results show that the recognition rate of this method is high,and the complexity is low.Partial occlusion of the image has a larger tolerance.(2)FPGA hardware design mainly includes Avalon bus interface,video input interface,display interface and algorithm of RTL interface module design,gives the hardware design related module of the RTL model,and compiled and verified,the algorithm module gives the simulation waveform of Verilog code are given as well the C++ code simulation results.(3)the integrity of the system to verify: receiving the input and output of the image,the experiments showed that the detection rate of the face recognition system reached 25 FPS,and made a lot of detection and the false alarm rate and relatively low rate.