基于人工神经网络的无线通信信号自动调制识别方法研究
刘学燕1
作者信息
- 1. 山东省邮电规划设计院有限公司 山东 济南 250101
- 折叠
摘要
对于一些复杂的调制方式,现有的自动调制识别方法可能无法完全准确地识别,现提出基于人工神经网络的无线通信信号自动调制识别方法.首先,基于人工神经网络构建无线通信信号调制模型,判定调制方式,并给出其对应的调制参数,然后,使用maxout函数作为人工神经网络卷积层和全连接层的激活函数,取代原先的ReLU函数,增加卷积层的数量,降低池化层的层数,提取无线通信信号特征参数,构建适合神经网络训练的特征矢量.最后,结合自适应调整的识别阈值,进行无线通信信号初始自动调制识别,通过相似度进一步优化调制识别方法,实现无线通信信号自动调制识别.通过测试数据集验证该方法的有效性和准确性.实验结果表明,使用人工神经网路模型在识别方面相较于文献方法更具优势,因此,可以证明基于人工神经网络的无线通信信号自动调制识别方法具有更高的精准度.
Abstract
For some complex modulation methods,the existing automatic modulation recognition methods may not be able to completely accurately identify the wireless communication signals.First,a wireless communication signal modulation model is built based on artificial neural network,the modulation mode is determined,and the corresponding modulation parameters are given.Then,maxout function is used as the activation function of convolutional layer and fully connected layer of artificial neural network,replacing the original ReLU function,increasing the number of convolutional layers and reducing the number of pooling layers.The feature parameters of wireless communication signal are extracted and the feature vector suitable for neural network training is constructed.Finally,the automatic modulation identification of wireless communication signals is carried out based on the adaptive adjustment of the recognition threshold,and the modulation identification method is further optimized by similarity to realize the automatic modulation identification of wireless communication signals.The validity and accuracy of the method are verified by the test data set.The experimental results show that the artificial neural network model is more advantageous than the literature method in recognition,it can be proved that the method of automatic modulation recognition of wireless communication signal based on artificial neural network has higher accuracy.
关键词
无线通信信号/人工神经网络/信号调制/自动识别Key words
wireless communication signal/artificial neural network/signal modulation/automatic recognition引用本文复制引用
出版年
2024
NETL
NSTL
万方数据