时频域特征融合网络雷达信号检测与识别OA
Radar signal detection and identification based on time-frequency domain features fusion network
针对传统雷达信号检测与识别在低信噪比环境下精度不足的问题,本文提出一种基于时频域特征融合的改进方法.以雷达信号同相及正交分量为输入,提取时域特征后,通过扩展离散傅里叶变换获取频域信息,构建融合通道空间注意力与长短时记忆网络的特征提取结构;利用交叉注意力实现时频域特征融合,采用多任务学习同步完成检测与脉内调制识别.实验表明,信噪比大于等于-6 dB时,识别率与检测概率均达99%;信噪比为-10 dB时,检测概率96.17%、识别准确率96.18%,均优于现有方法,有效提升复杂环境下的检测识别性能.
To solve the low accuracy of traditional radar signal detection and recognition in low signal-to-noise ratio environments,this paper proposes an improved method.With the in-phase and quadrature components of radar sig-nals as input,after extracting time-domain features,frequency-domain information is obtained via extended dis-crete Fourier transform,and a feature extraction structure combining channel spatial attention and long short-term memory network is constructed.Cross-attention is used for time-frequency feature fusion,and multi-task learning is adopted to simultaneously complete detection and intra-pulse modulation recognition.Experiments show that at SNR≥-6 dB,both recognition rate and detection probability reach 99%;at-10 dB,the detection probability is 96.17%and recognition accuracy is 96.18%,both superior to existing methods,effectively improving detection and recognition performance in complex environments.
肖易寒;李本正;秦长海
哈尔滨工程大学 信息与通信工程学院,黑龙江 哈尔滨 150001||哈尔滨工程大学 先进船舶通信与信息技术工业和信息化部重点实验室,黑龙江 哈尔滨 150001哈尔滨工程大学 信息与通信工程学院,黑龙江 哈尔滨 150001||哈尔滨工程大学 先进船舶通信与信息技术工业和信息化部重点实验室,黑龙江 哈尔滨 150001中国船舶集团有限公司第七二三研究所,江苏 扬州 225000||上海交通大学 计算机学院,上海 200240
信息技术与安全科学
雷达信号检测雷达调制识别时频域特征融合扩展离散傅里叶变换特征提取通道空间注意力交叉注意力多任务学习深度学习
radar signal detectionradar modulation recognitiontime-frequency domain feature fusionextended discrete Fourier transformfeature extractionconvolutional block attention modulecross-attentionmulti-task learningdeep learning
《哈尔滨工程大学学报》 2026 (5)
1136-1145,10
评论