中文摘要：

特征选择是模式识别领域中的一个重要问题，其本质是一个寻优的过程。在特征选择算法FOS的基础上，提出了一种特征选择算法。该算法选择出了最能代表样本大多数特征的特征，构成有效特征子集，实现了数据的降维。基于南京理工大学NUST603HW手写汉字库的实验结果表明，该算法不仅提高了识别率，而且稳定性更强。

英文摘要：

Feature selection is an important problem in the fields ofpattem recognition, its essence is an optimization process. A new feature selection algorithm based on FOS is proposed. The new algorithm selected the features, which represent the most features in the datasets, formed effective feature subsets, and reduced the dimensions ofdatasets in the recognizing process. The experiment is tested on Nanjing University of Science and Technology NUST603HW handwritten Chinese character database, the results based on the new algorithm indicated that the recognition rate is improved and the classification results are more robust.