基于动态粒子群优化信息熵的人脸识别方法  

An Information Entropy Method Based on Dynamic Particle Swarm Optimization for Face Recognition

作　　者：白杨[1] 

机构地区：[1]温州大学电子信息系,浙江温州325035

出　　处：《计算机仿真》2008年第7期174-176,272,共4页Computer Simulation

基　　金：浙江省自然科学基金(Y104592);浙江省教育厅科研项目(20041032)

摘　　要：人脸识别研究的目标主要有两个,一是提高识别正确率,二是降低训练与识别时间。信息熵等方法主要取决于参数选择,然而传统的优化算法难以解决此问题。粒子群算法等智能搜索技术可在较少的时间内给出问题的近似解。动态粒子群优化算法是在经典的微粒群算法的基础上所提出的一种高效的收敛性、稳定性的进化算法。采用动态粒子群算法对信息熵优化寻找最优参数,并结合特征提取方法,用于人脸图像的识别中,为人脸识别问题的研究开辟了新的途径。最后通过仿真实验得出结论表明,既减少了计算复杂度,降低训练与识别时间,又保证实时性,提高识别正确率,得到了理想的结果。There are two targets in face recognition： enhancing recognition correctness and reducing training and recognition time.Information Entropy Method（IEM）has proved to be a powerful technique for solving problems in pattern classification and regression,but its capacity mainly depends on the parameter selection.Parameter selection for IEM is very complex in nature and quite hard to solve by conventional optimization techniques.The intelligent search technologies,such as particle swarm optimization algorithm,can give a similar solvent of problems with less time.Dynamic particle swarm optimization（DPSO） on the basis of classical particle swarm optimization is a better method with convergence and stability.Information Entropy Method trained with DPSO algorithm which is simple and rapid,integrated with feature extraction,is applied in the recognition of face images,and it is a new way for face recognition.Finally through simulation some relational conclusions are obtained.Experimental results show that the new method can not only obtain ideal face image recognition results but also reduce the calculation complexity,cut down the training and recognition time and achieve real time recognition,and it can enhance the recognition correctness.

关 键 词：人脸识别 粒子群算法 信息熵 动态粒子群算法 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]