Scholarlink

<p>The underwater acoustic channel has serious multipath effects, frequency selective fading and fast time-varying channel structure, which lead to distortion and attenuation of the received signal. In view of the high bit error rate of the traditional demodulation method and its inability to adapt to the rapid changes of the underwater acoustic channel, this paper introduces a novel underwater acoustic receiver architecture that integrates an enhanced Convolutional Neural Network (CNN) with a Bidirectional Long Short-Term Memory (BiLSTM) model utilizing Orthogonal Frequency Division Multiplexing (OFDM). The powerful feature extraction capability of the convolutional neural network (CNN) is used to extract channel features, followed by a deconvolution layer, and the idea of image super-resolution is used for channel estimation to combat frequency selective fading. At the same time, a bidirectional long short-term memory network (BiLSTM) is used to capture the timing information in the signal, learn the long-term dependency of the signal, and restore and reconstruct the received signal. Simulation and experimental results show that the receiver based on CNN+BiLSTM proposed in this paper can achieve a gain of up to 5dB compared with the traditional method, and has a lower bit error rate and is more robust than other deep learning methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

OFDM underwater acoustic communication receiver based on improved CNN and BiLSTM

  • 期刊论文

  • 发表日期：

    2026年6月25日

查看全文
  • Shuhiu Zhang,
  • Xinghai Yang,
  • Zihe Ren,
  • Jingjing Wang

摘要

The underwater acoustic channel has serious multipath effects, frequency selective fading and fast time-varying channel structure, which lead to distortion and attenuation of the received signal. In view of the high bit error rate of the traditional demodulation method and its inability to adapt to the rapid changes of the underwater acoustic channel, this paper introduces a novel underwater acoustic receiver architecture that integrates an enhanced Convolutional Neural Network (CNN) with a Bidirectional Long Short-Term Memory (BiLSTM) model utilizing Orthogonal Frequency Division Multiplexing (OFDM). The powerful feature extraction capability of the convolutional neural network (CNN) is used to extract channel features, followed by a deconvolution layer, and the idea of image super-resolution is used for channel estimation to combat frequency selective fading. At the same time, a bidirectional long short-term memory network (BiLSTM) is used to capture the timing information in the signal, learn the long-term dependency of the signal, and restore and reconstruct the received signal. Simulation and experimental results show that the receiver based on CNN+BiLSTM proposed in this paper can achieve a gain of up to 5dB compared with the traditional method, and has a lower bit error rate and is more robust than other deep learning methods.