基于加权矢量神经网络的辐射源识别方法

现有的很多分类识别方法包括基于专家系统的方法1]、基于贝叶斯理论的方法、基于模糊模式识别的方法2]、基于最近邻的方法3]、基于人工神经网络的方法3]等等在辐射源识别中都有比较成功的应用,但这些方法一般都针对测量参数为标量形式的测量值进行处理,在一定程度上解决了由于参数测量误差所引起的辐射源识别问题,对于误差的另一种情形,即测量参数为区间类型模糊值的情况讨论却较少,文献提出了一种基于模糊IF-Then规则的神经网络算法,给出了能够处理模糊输入的神经网络体系结构,同时给出了一种基于代价函数的学习算法,其代价函数由实际模糊输出和无模糊输出决定,通过学习该网络能够实现模糊输入到模糊输出的非线性映射。

Emitter Recognition Based on the Weighted Vector Neural Network

Many existing classification methods include methods based on expert system,Bayesian theory,fuzzy pattern recognition method,the nearest neighbor method,artificial neural networks,etc.are successfully used in the emitter identification,but these methods are generally aimed at measuring parameters for the scalar form of measurement processing,to some extent solved the parameter measurement error caused by emitter identification problem,in another case,measurement parameter values for the interval type of fuzzy situation has been less discussed.The literature presents neural network algorithm based on fuzzy IF-Then rules that can deal with fuzzy input neural network architecture,while presents learning algorithm based on cost function,and the cost function is determined by the actual fuzzy output and non-fuzzy output,the network can achieve the nonlinear mapping of fuzzy input to fuzzy output.