基于典型样本的信度函数分配的构造方法

摘要

D-S证据理论信度函数分配的取值是得到较为准确的融合结果的关键，然而传统方法如采用隶属度函数、正态分布等得到的信度函数分配都具有较大的主观性。为使信度函数分配更具客观性，在总结其它方法的基础上，提出了基于典型样本的信度函数分配构造方法。首先采集各目标模式下的样本，并判断每一模式下的各条证据服从何种概率分布，利用相应的概率公式计算待识别目标模式的各条证据的概率密度，然后进行归一化处理，最后利用联合规则得到融合结果。实例表明利用此法可得到较为准确的融合结果，不仅提高了判别结果的准确性，而且降低了不确定度，并再一次证明了融合诊断结果比单一数据具有更高的可靠性。%The value of D-S evidence theory of belief function assignment is the key to get accurate fusion results, while traditional methods, such as the belief function assignment got through the usage of the membership function and the normal distribution, tend to be more subjective.In order to make the belief function assignment more ob-jective.This paper puts forward a method constructing belief function assignment after combining other methods and analyzing the typical samples.First, samples were collected for each target mode, and which probability distribu-tions the evidence of each mode obeys was judged.Then, the probability density of each piece of evidence was cal-culated by using corresponding probability formula.Finally, normalized processing was carried on, and the union rule was used to obtain the fusion results.The example shows that using this method can get more accurate results of fusion, and this method can not only improves the accuracy of the result of discrimination, but also reduce the uncertainty.In addition, it proves that the fusion diagnosis has a higher reliability than a single data.