Development Of Face Recognition System Based On Improved PCA And SVM For Android Platform

With the rapid development of artificial intelligence,the theory and technology of face recognition are gradually moving from laboratory to engineering application.Based on the mobile phone of Android operating system,a face recognition system based on improved Principal Component Analysis(short PCA)and Support vector machine(short SVM)is studied in this paper.The paper further studies the shortcomings of the Proximal SVM(short PSVM)algorithm,and proposes an improvement method.The system covers the main procedure of the typical face recognition system,including: capturing raw face image,standardizing face image,constructing the local information database including face sample dataset,extracting face feature,classifying,displaying the final results.The main research contents of this paper are as follows.(1)Mobile face recognition system is designed with face image capturing and preprocessing.According to the characteristics of the mobile phone,this paper designs the system architecture,the overall operating flow of the system,face image capturing and standardizing operation,etc.The theoretical basis and optimization of image preprocessing such as geometric standardization,graying,noise filtering and histogram equalization are analyzed and studied.(2)Facial feature extraction algorithm based on improved PCA is researched and implemented.This paper analyzes and studies the important PCA and its improved algorithm in face feature extraction field,compares and verifies the performance of these methods through a series of experiments,studies the realization of feature extraction algorithm on mobile phone,and obtains the face feature extraction transformation matrix.(3)Face classifier algorithm based on improved SVM is researched and implemented.The SVM and its improved algorithm are studied and analyzed,and the problem of decreasing dimensions of sample matrix is considered for the Proximal SVM in the process of nonlinear samples.An improved scheme based on sample selection is proposed to settle this disadvantage of PSVM.The relative experiments are designed to compare and verify the perfermence of the improved method.The implementation of classifier algorithm also is studied.(4)System function realization and test are conducted.According to the previous theoretical research,overall analysis,design and the specific functions of the system are implemented.To make the program run faster,the paper researchs and uses the JNI(namely Java Native Interface)and Android local development technology based on NDK(namely Native Development Kit).In order to manage the information of personnel and face samples needed in the process of identification,the local database is built,the module of result processing is designed for better interact for users,and the system’s operation and perfermence are tested to verify the integrity and feasibility of the system.