前导检测是水声通信中的关键步骤。传统信号检测算法无法有效克服多径和干扰的影响，在复杂水声信道环境下，检测性能会明显下降。卷积神经网络具有强大的表征学习能力，能够通过大量数据自动提取图像特征，在图像识别应用中具有突出优势。以双曲调频信号为研究对象，采用时频分析技术同时获取信号时域和频域信息，得到时频谱图，并通过均值滤波处理提高信噪比。将时频谱图作为卷积神经网络的输入，训练并测试网络，最终得到HFM信号的检测结果。千岛湖试验结果表明，该方法能够大幅提升远距离、强多径条件下前导信号的检测概率，与传统检测算法相比性能优越。
Preamble detection is an important step in underwater acoustic communications. The traditional detection methods cannot effectively overcome the multipath and interference, so the detection performance will be significantly decline when applied in a complex channel. Convolutional neural network (CNN) has outstanding advantages in the image recognition because of its strong learning ability and automatic feature extraction. Take the HFM signal as the research object, use the time-frequency analysis technology to get the information of time and frequency simultaneously, then obtain the time-frequency spectrum. The mean- filtering is carried out to reduce the image noise. Take the spectrum as the input of CNN, train and test the network, then obtain detection result. The experimental results in Thousand island lake show that the detection method based on CNN and STFT can greatly improve the detection probability in long distance and strong multipath channel, and has superior performance compared with the traditional methods.
2023,45(22): 138-142 收稿日期：2022-11-3
DOI：10.3404/j.issn.1672-7649.2023.22.026
分类号：U666.7
基金项目：国家自然科学基金资助项目（62192711）；中国科学院声学研究所自主部署“目标导向”类项目资助项目（MBDX202104）
作者简介：张震(1989-),男,硕士,副研究员,研究方向为水声信号处理。
参考文献：
[1] TAN Q Y, WANG Y Y, MA X L. Preamble detection based on cyclic features of zadoff-chu sequences for underwater acoustic communications[J]. IEEE Signal Processing, Lett, 2019, 26(8).
[2] LI W, ZHOU S L, Peter Willett, et al. Preamble detection for underwater acoustic communications based on sparse channel identification[J]. IEEE Journal of Oceanic Engineering, 2019, 44(1).
[3] QIAO G, ZEESHAN Babar, MA L, et al. MIMO-OFDM underwater acoustic communication systems-A review[J]. Physical Communication, 2017, 23.
[4] 赵方真. 水声通信系统中的前导信号检测与识别性能分析[D]. 哈尔滨: 哈尔滨工业大学, 2020.
[5] 王大宇, 王志欣, 张光普. 基于改进谱减算法的水声通信信号检测方法[J]. 应用科技, 2020, 47(3): 69–73
WANG Dayu, WANG Zhixin, ZHANG Guangpu. Detection method of underwater acoustic communications signal based on improved spectral subtraction algorithm[J]. Applied Science and Technology, 2020, 47(3): 69–73
[6] 叶辉. 宽带频谱检测技术研究以及FPGA实现[D]. 成都: 电子科技大学, 2020.
[7] 孙伟, 彭华, 李天昀, 等. 基于时频图像处理的宽带特定信号检测方法[J]. 信息工程大学学报, 2020, 21(5): 545–551
SUN Wei, PENG Hua, LI Tianyun, et al. Broadband specific signal detection method based on time-frequency image processing[J]. Journal of Information Engineering University, 2020, 21(5): 545–551
[8] Xl A, TO B, MK A, et al. An effective method for laboratory acoustic emission detection and location using template matching[J]. Journal of Rock Mechanics and Geotechnical Engineering, 2022.
[9] 袁冠杰, 王虹, 陈静, 等. 基于特征提取融合判决的卫星干扰识别算法[J]. 微波学报, 2021, 37(S1): 241–244
YUAN Guanjie, WANG Hong, CHEN Jing, et al. Satellite jamming recognition algorithm based on feature extraction and fusion decision[J]. Journal of Microwaves, 2021, 37(S1): 241–244
[10] S. Raghu, Natarajan Sriraam, Yasin Temel, et al. Performance evaluation of DWT based sigmoid entropy in time and frequency domains for automated detection of epileptic seizures using SVM classifier[J]. Computers in Biology and Medicine, 2019, 110: 127-143.
[11] 魏煜宁, 张劲东, 李勇, 等. 雷达干扰信号识别决策树的自动化设计方法[J]. 电光与控制, 2020, 27(4): 82–86.
WEI Yuning, ZHANG Jingdong, LI Yong, et al. A method for automatic design of decision tree in radar jamming signal recognition[J]. Electronics Optics & Control, 2020, 27(4): 82–86.
