通信学报 ›› 2019, Vol. 40 ›› Issue (10): 119-126.doi: 10.11959/j.issn.1000-436x.2019181

• 学术论文 • 上一篇    下一篇

基于降噪自编码器的水声信号增强研究

殷敬伟1, 罗五雄1, 李理1, 韩笑1, 郭龙祥1, 王建峰2   

  1. 1 哈尔滨工程大学水声工程学院,黑龙江 哈尔滨 150001
    2 中国船舶工业综合技术经济研究院,北京 100081
  • 修回日期:2019-06-24 出版日期:2019-10-25 发布日期:2019-11-07
  • 作者简介:殷敬伟(1980- ),男,黑龙江尚志人,博士,哈尔滨工程大学教授、博士生导师,主要研究方向为水声通信及信号处理。|罗五雄(1995- ),男,湖北汉川人,哈尔滨工程大学硕士生,主要研究方向为水声信号处理、机器学习。|李理(1987- ),男,黑龙江哈尔滨人,博士,哈尔滨工程大学讲师,主要研究方向为图像及语音信号处理、机器学习、模式识别。|韩笑(1988- ),男,山东兖州人,博士,哈尔滨工程大学讲师、硕士生导师,主要研究方向为水声通信侦察与对抗、极地声学、声纳信号处理。|郭龙祥(1976- ),男,黑龙江哈尔滨人,博士,哈尔滨工程大学副教授、硕士生导师,主要研究方向为水声信号处理。|王建峰(1984- ),男,河北沧州人,博士,中国船舶工业综合技术经济研究院工程师,主要研究方向为数据分析、挖掘与计算机技术。
  • 基金资助:
    国家自然科学基金资助项目(51779061);霍英东教育基金资助项目(151007);黑龙江省杰出青年科学基金资助项目(JC2017017)

Enhancement of underwater acoustic signal based on denoising automatic-encoder

Jingwei YIN1, Wuxiong LUO1, Li LI1, Xiao HAN1, Longxiang GUO1, Jianfeng WANG2   

  1. 1 College of Underwater Acoustic Engineering,Harbin Engineering University,Harbin 150001,China
    2 China Institute of Marine Technology and Economy,Beijing 100081,China
  • Revised:2019-06-24 Online:2019-10-25 Published:2019-11-07
  • Supported by:
    The National Natural Science Foundation of China(51779061);The Fok Ying-Tong Education Foundation of China(151007);The Natural Science Funds for Distinguished Young Scholar of Heilongjiang Province(JC2017017)

摘要:

针对主动声呐中回波信号特征提取困难的问题,提出了一种利用降噪自编码器与卷积降噪自编码器相结合的自编码器算法。首先利用降噪自编码器在信号整体上的降噪优势,对含噪信号进行预处理;然后结合卷积降噪自编码器对信号局部特征的优化,对信号进行局部降噪,从而实现信号增强。所提算法直接以接收信号的时域波形作为特征输入,保留了信号的幅度与相位特征。实验结果表明,所提算法不仅有效降低了信号中的噪声分量,而且在时域和频域上均达到了较好的恢复效果。

关键词: 线性调频信号, 降噪自编码器, 卷积降噪自编码器, 信号增强

Abstract:

Aiming at the difficulty of feature extraction of echo signal in active sonar,a self-encoder algorithm based on the combination of denoising self-encoder and convolution denoising self-encoder was proposed.Firstly,the preprocessing of noisy signal was carried out by using the advantage of denoising self-encoder in signal as a whole,and then the local feature of signal was optimized by combining convolutional denoising self-encoder to denoise the signal locally,so as to enhance the signal.The time domain waveform of the received signal is used as the feature input by the algorithm,and retains the signal’s amplitude and phase characteristics.The experimental results show that the algorithm not only effectively reduces the noise component in the signal,but also achieves better recovery effect in both time and frequency domains.

Key words: LFM, denoising automatic-encoder, convolution denoising automatic-encoder, signal enhancement

中图分类号: 

No Suggested Reading articles found!