Research On Face Recognition Based On Novel Bacterial Foraging Algorithm And Nonlinear Statistical Dimension Reduction

In many biometric identification methods,,face recognition,which has the characteristics of direct viewing,non contact and parallelism,has been successfully applied to many fields,including statistics is the core of dimension reduction has been the focus of researchers parties.At present,the most commonly used method of dimension reduction,principal component analysis(PCA),is to determine the dimension of matrix reduction,which is the number of feature vectors.However,in this case,it is easy to ignore the feature vectors with small eigenvalues and high classification information,which will affect the recognition efficiency.Therefore,this paper proposes an improved method based on a novel swarm intelligence optimization algorithm,the bacterial foraging algorithm(BFO)and the nonlinear dimension reduction algorithm.In addition,the standard bacterial foraging algorithm is improved and applied to face recognition.The main contents of researching and works are as follows:(1)For dimension reduction algorithm in the process of feature extraction part only rely on the limitations of cumulative contribution rate to select feature vector,this paper presents a nonlinear dimension reduction algorithm and bacterial foraging algorithm(BFO)method.The dimension reduction algorithm to extract the feature vector as the initial position of the bacteria in the search space,according to the fitness function value of the feature vector for iterative search,extraction of near optimal feature vector has a greater contribution to classification,to achieve dimension reduction optimization objective.(2)In the standard BFO algorithm,the chemotactic operation of the bacteria in the swimming step to the fixed value,and the mechanism of self replication breeding the next generation of bacteria,this situation leads to the diversity of the population decreased and the slow convergence of the algorithm and other issues.In order to solve these problems,this paper proposes an adaptive method to adjust the step size;For copy,copy of adaptive probability of crossover method is proposed in this paper,the algorithm as much as possible,jump out of local minimum value,improve the convergence speed.In order to further prove the efficiency and feasibility of bacterial foraging algorithm,in this paper,the nonlinear algorithm and the algorithm convergence and comparative test is applied to face recognition.