摘要
针对传统特征提取的个体识别算法难以满足实际电磁环境条件下对同型号个体进行识别的实际需求的问题,设计了一种基于多隐层神经网络的通信辐射源个体识别算法。首先对实际采集信号进行预处理,将预处理后的信号制作为适合训练的数据集。然后通过制作的大量数据进行网络训练,结合优化策略对网络进行优化,并通过测试数据结合softmax层和反向传播的方法对网络进行迭代校准,得到针对辐射源个体识别的网络。通过实际采集信号对网络进行训练和测试以验证其有效性。实验得到平均分类识别准确率为90%,能够取得较好效果。
Aiming at the probelm that indicidual identification based on traditional feature extraction method cannot meet the actual needs of identifying individuals of the same type under the actual electromagnetic environment,a individual identification algorithm based on multi-hidden layer neural network was proposed.Firstly,preprocess the actual collected signals and make it into a suitable data sets for neural network training.Then,the neural network was trained through a large amount of data and optimized by combining optimization strategies.Calibrated by test data with softmax layer and back propagation method to obtain a netwoek for individual identification.The network was trained and tested by actually collecting signals to verify its effectiveness.Experiments show that the average accuracy rate is 90%.
作者
耿梦婕
张君毅
Geng Mengjie;Zhang Junyi(The 54 Rearch Institute of CETC Shijiazhuang 050081,China)
出处
《电子测量技术》
2019年第21期137-142,共6页
Electronic Measurement Technology
关键词
辐射源个体识别
多隐层神经网络
数据集
网络优化
indicidual identification
multi-hidden layer neural network
data sets
neural network calibrated
