针对目前多源信息融合存在的问题,本文提出采用模拟人类思维的方法来实现信息融合。该方法首先利用粗集理论的强定性分析能力对多源输入信息进行约简,消除了样本中的噪声和冗余,在此基础上,再利用支持向量机对约简信息进行融合。为了获得最优的融合精度,该方法还利用遗传算法对融合进行了优化,文字识别的实例研究表明,该方法具有良好的容错性、稳健性和准确性。 Aiming at the existing problems of multi-source information fusion, this paper proposes the method of simulating human thinking to realize the information fusion. Firstly, the method reduces the multi-source input information by using the strong qualitative analysis ability of rough set theory and eliminates the noise and redundancy in the sample. Based on this, the support vector machine is used to fuse the reduction information. In order to obtain the best fusion accuracy, the method also uses genetic algorithm to optimize the fusion. The case study of word recognition shows that the method has good fault tolerance, robustness and accuracy.