基于SVM的多特征数据的概率融合方法

Proability fusion method for the data from different feature based on support vector machine

针对刻画研究对象的独立多特征的多来源数据或信息,目的为对总体类别识别问题,研究了利用支持向量机方法对各特征分别进行模式识别后,引入样本的后验概率,并以此形成似然函数,从而建立多源数据特征级概率融合方法,该方法具有不必对数据进行预处理,处理的数据可以是不同质、高维、小样本数据,没有强的先验假设,能用机器自动实现等特点。

In the procedure of independent multiple feature data or information fusion, a probability method based on the maximum like lihood criterion, which constructed by introducing posteriori probability into support vector machines, has been proposed for the objective of data fusion is decision the type of totality. The method have character of that data have not needed pro -processing and may be different feature, high dimension, small seize, it have no prior assumption and can be automatic realization by machines.