Abstract: In order to enhance the palmprint recognition performance, it proposes a novel palmprint recognition model based on simultaneously selecting features and classifier parameters according to relation between the dimensions of the Principal Component Analysis（PCA） and parameters of Support Vector Machines（SVM）. The palmprint image is preprocessed, and then the dimensions of PCA and parameters of SVM are taken as a particle, the optimal palmprint features and parameters of SVM are obtained simultaneously by information exchange and cooperation of particle swarms, the optimal palmprint recognition model is established based on the selected dimensions and parameters, the performance of model is tested by Po1yU palmprint data. The results show that the proposed model can obtain recognition rates of the palmprint 94%, the prediction results are significantly better than reference models which features and classifier parameters are selected separately.

Key words: palmprint recognition, support vector machine, principal component analysis, simultaneously selection, particle swarm optimization algorithm

摘要： 为了进一步提高掌纹识别系统性能，充分利用主成分分析特征维数和支持向量机参数之间的联系，提出了一种特征维数和分类器参数统一选择的掌纹识别模型（Features-Classifier）。对掌纹图像进行预处理，将主成分分析图像特征维数和支持向量机参数作为一个粒子，在统一的目标函数下通过粒子之间的信息交流和相互协作，找到最优掌纹特征和分类器参数，根据最优掌纹特征和分类器参数建立掌纹图像识别模型，并采用Po1yU掌纹数据库对模型性能进行仿真实验。结果表明，Features-Classifier的掌纹平均识别率达到94%以上，识别结果明显优于独立、分开选择特征维数和分类器参数的掌纹识别模型。

关键词: 掌纹识别, 支持向量机, 主成分分析, 统一选择, 粒子群优化算法