全文下载： 
202304016.pdf
文章编号： 1672-6987（2023）04-0120-07； DOI： 10.16351/j.1672-6987.2023.04.016
吴承安， 赵文龙， 王景景*(青岛科技大学 信息科学技术学院，山东 青岛 266061）
摘要： 复杂的水下信道环境和严重的噪声干扰给接收机正确识别调制方式带来巨大挑战。为解决该问题，结合门控循环单元(gated recurrent unit，GRU)和残差神经网络(residual network，ResNet)的自动特征提取和学习能力，设计了一种基于混合神经网络的时空特征融合模型，称为门控和残差融合模型(GRU and ResNet fused model，G&RFM)。与传统的调制识别技术相比，该方法无需人工提取特征，就能获得较高的识别精度。实验结果表明，所提出的G&RFM在南海数据集上的验证准确率为9831%。与改进前的网络模型相比，本工作所提出的G&RFM具有更高的识别精度，有效提升了神经网络学习性能。
关键词： 水声信号； 调制识别； 混合神经网络； 特征融合
中图分类号： TN 929.3文献标志码： A
引用格式： 吴承安， 赵文龙， 王景景. 基于时空特征融合的水声信号调制识别［J］. 青岛科技大学学报(自然科学版), 2023, 44(4): 120-126.
WU Cheng′an, ZHAO Wenlong, WANG Jingjing. Modulation recognition of underwater acoustic signals based on fusion of spatiotemporal feature［J］. Journal of Qingdao University of Science and Technology（Natural Science Edition), 2023， 44（4）： 120-126.
Modulation Recognition of Underwater Acoustic Signals
Based on Fusion of Spatiotemporal Feature
WU Cheng′an，ZHAO Wenlong，WANG Jingjing
(College of Information Science and Technology, Qingdao University of Science and Technology, Qingdao 266061, China)
Abstract: It is a huge challenge for the receiver to correctly identify the modulation types due to the complex underwater channel environment and severe noise interference. To solve this problem, combined with the automatic feature extraction and learning ability of Gated Recurrent Unit (GRU) and Residual Network (ResNet), a spatiotemporal feature fusion model based on hybrid neural network is designed, which is called GRU and ResNet Fused Model (G&RFM). As compared with the traditional modulation recognition techniques, this method achieves higher recognition accuracy without manual feature extraction. The experimental results show that the validation accuracy of the proposed G&RFM on the South China Sea data set is 9831%. Compared with the unimproved network model, the proposed G&RFM has a higher recognition accuracy and effectively improves the learning performance of neural network.
Key words: underwater acoustic signal; modulation recognition; hybrid neural network; feature fusion
收稿日期： 2022-07-28
基金项目： 国家自然科学基金联合基金重点项目（U1806201）；山东省自然科学基金重大基础研究项目（ZR2021ZD12）.
作者简介： 吴承安(1996—),男,硕士研究生.*通信联系人.
