基于混合注意力原型网络的调制识别算法

Q3 Engineering

西北工业大学学报 Pub Date : 2022-12-01 DOI:10.1051/jnwpu/20224061375

Yi Pang, Hua Xu, Lei Jiang, Yunhao Shi, Xiang Peng

{"title":"基于混合注意力原型网络的调制识别算法","authors":"Yi Pang, Hua Xu, Lei Jiang, Yunhao Shi, Xiang Peng","doi":"10.1051/jnwpu/20224061375","DOIUrl":null,"url":null,"abstract":"针对极少量带标签样本条件下的通信信号调制识别难题, 提出一种基于混合注意力原型网络的调制识别算法。综合元学习和度量学习的思想, 在原型网络框架下通过特征提取模块将信号映射至一个新的特征度量空间, 并通过比较该空间内各类原型与查询信号之间的距离确定查询信号调制样式。根据通信信号IQ分量的时序特点设计了由卷积神经网络和长短时记忆网络级联的特征提取模块, 并引入卷积注意力机制提升关键特征的权重; 采用基于Episode的训练策略, 使算法可泛化到新的信号识别任务中。仿真结果表明, 所提算法在每类信号只有5个带标签样本(5-way 5-shot)时平均识别率可达85.68%。","PeriodicalId":39691,"journal":{"name":"西北工业大学学报","volume":" ","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Modulation recognition algorithm based on mixed attention prototype network\",\"authors\":\"Yi Pang, Hua Xu, Lei Jiang, Yunhao Shi, Xiang Peng\",\"doi\":\"10.1051/jnwpu/20224061375\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"针对极少量带标签样本条件下的通信信号调制识别难题, 提出一种基于混合注意力原型网络的调制识别算法。综合元学习和度量学习的思想, 在原型网络框架下通过特征提取模块将信号映射至一个新的特征度量空间, 并通过比较该空间内各类原型与查询信号之间的距离确定查询信号调制样式。根据通信信号IQ分量的时序特点设计了由卷积神经网络和长短时记忆网络级联的特征提取模块, 并引入卷积注意力机制提升关键特征的权重; 采用基于Episode的训练策略, 使算法可泛化到新的信号识别任务中。仿真结果表明, 所提算法在每类信号只有5个带标签样本(5-way 5-shot)时平均识别率可达85.68%。\",\"PeriodicalId\":39691,\"journal\":{\"name\":\"西北工业大学学报\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2022-12-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"西北工业大学学报\",\"FirstCategoryId\":\"1093\",\"ListUrlMain\":\"https://doi.org/10.1051/jnwpu/20224061375\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"Engineering\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"西北工业大学学报","FirstCategoryId":"1093","ListUrlMain":"https://doi.org/10.1051/jnwpu/20224061375","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"Engineering","Score":null,"Total":0}

引用次数: 0

摘要

A modulation recognition algorithm based on a hybrid attention prototype network is proposed to address the modulation recognition problem of communication signals with very few labeled samples. Integrating the ideas of meta learning and metric learning, the signal is mapped to a new feature metric space through a feature extraction module within the prototype network framework, and the modulation style of the query signal is determined by comparing the distances between various prototypes and the query signal in this space. A feature extraction module was designed based on the temporal characteristics of the communication signal IQ component, which is cascaded by a convolutional neural network and a long and short term memory network. The convolutional attention mechanism was introduced to enhance the weight of key features; Adopting an Episode based training strategy, the algorithm can be generalized to new signal recognition tasks. The simulation results show that the proposed algorithm has an average recognition rate of 85.68% when there are only 5 labeled samples (5-way 5-shot) for each type of signal.

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

Modulation recognition algorithm based on mixed attention prototype network

针对极少量带标签样本条件下的通信信号调制识别难题, 提出一种基于混合注意力原型网络的调制识别算法。综合元学习和度量学习的思想, 在原型网络框架下通过特征提取模块将信号映射至一个新的特征度量空间, 并通过比较该空间内各类原型与查询信号之间的距离确定查询信号调制样式。根据通信信号IQ分量的时序特点设计了由卷积神经网络和长短时记忆网络级联的特征提取模块, 并引入卷积注意力机制提升关键特征的权重; 采用基于Episode的训练策略, 使算法可泛化到新的信号识别任务中。仿真结果表明, 所提算法在每类信号只有5个带标签样本(5-way 5-shot)时平均识别率可达85.68%。

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

西北工业大学学报 Engineering-Engineering (all)

CiteScore

1.30

自引率

0.00%

发文量

6201

审稿时长

12 weeks

期刊介绍：