基于改进SOFM网络的调制方式的自动识别

提出了应用改进的自组织特征映射神经网络自动识别信道调制方式的方法．首先，提取信号特征矢量；然后对网络的输出结构进行优化，引入反馈策略，并采用自组织分析方法，对训练后的节点进行动态聚类．最后，利用训练好的网络对加载不同噪声的调制信号进行识别．仿真结果表明，该识别法识别率高于传统的决策论法的识别率，平均值高3．2个百分点左右．

Automatic recognition of modulation types based on improved SOFM NN

An automatic recognition method of the channel modulation types by is suggested using improved self-organizing feature map neural network(SOFM NN).The signal feature vector is extracted at first,and then the network output structure is optimized.Feedback strategy is introduced into the training of SOFM NN.Then self-organizing feature analysis is used to cluster dynamically after training.Finally,the trained network recognizes signals with different noises.Experimental results indicate that the recognition rate is 3.2 percent higher than neural network recognition method with conventional technology.