基于双曲正切特征和双特征融合神经网络的调制方式识别

删除或更新信息，请邮件至freekaoyan#163.com(#换成@)

本站小编 Free考研考试/2024-01-16

-->

张兆军,高银锐,邱天爽,李芃芃,栾声扬,陈薇.基于双曲正切特征和双特征融合神经网络的调制方式识别[J].,2023,63(1):93-101

基于双曲正切特征和双特征融合神经网络的调制方式识别

Automatic modulation classification based on hyperbolic-tangent-based patterns and dual-feature-fusion neural network

DOI：10.7511/dllgxb202301012

中文关键词:调制方式识别混合噪声双曲正切函数注意力机制特征融合

英文关键词:automatic modulation classification mixed noise hyperbolic tangent function attention mechanism feature fusion

基金项目:2022江苏省青蓝工程项目；徐州市科技计划项目(KC22290KC21001)；国家自然科学基金资助项目(61801197)；江苏省自然科学基金资助项目(BK20181004)；江苏省高等学校基础科学(自然科学)研究项目(21KJB52000518KJB510012).

作者	单位
张兆军,高银锐,邱天爽,李芃芃,栾声扬,陈薇

摘要点击次数:190

全文下载次数:231

中文摘要:

由于调制方式识别能够在先验知识不足的情况下判断接收信号的调制类型，故而在各类无线通信系统中起到重要作用，许多****也围绕该问题进行了深入的研究，并在高斯噪声条件下取得了诸多进展．但是，由于自然或人为因素，有时候噪声中会出现尖峰，导致其具有脉冲性．此时，就无法继续采用高斯模型对其进行描述，因为相关的方法会发生严重的性能退化或者完全失效．为了解决该问题，首先借助双曲正切函数提出了两类改进的特征；然后基于注意力机制和门控循环单元等设计了一种双特征融合的深度神经网络作为分类器；最后，在模式识别框架下提出了一种新的调制方式识别方法．实验结果表明，所提方法能够有效抑制高斯和非高斯混合噪声，同时取得良好的调制方式识别结果．

英文摘要:

Automatic modulation classification could recognize the modulation type of received signal when the prior knowledge is not enough. Therefore, it plays a critical role in many wireless communication systems. Besides, many scholars have focused on this topic and made solid approaches under the Gaussian noise assumption. However, due to natural or human factors, sharp spikes may occur in the noise, resulting in impulsiveness. In this scenario, it is no longer appropriate to describe the characteristics of the noise through Gaussian model since related methods may experience severe performance deterioration or total failure. To cope with this problem, two novel patterns modified by the hyperbolic tangent function are first proposed. Then, a dual-feature-fusion deep neural network is designed as the classifier based on the attention mechanism and gated recurrent unit. In the end, a novel automatic modulation classification method is presented in a manner of pattern recognition. Experiments validate its robustness to the Gaussian and non-Gaussian mixed noise and its superior modulation classification capability.

查看全文查看/发表评论下载PDF阅读器

关闭

基于双曲正切特征和双特征融合神经网络的调制方式识别

本站小编 Free考研考试/2024-01-16

相关话题/

领限时大额优惠券,享本站正版考研考试资料!