无线电工程

水下通信信号模式识别是非合作水下通信信号识别中的关键一步，然而，水声信道的复杂多变给水下通信信号模式识别带来了很大挑战。针对传统算法模型复杂度高、提取特征多的问题，提出了一种残差网络和时序卷积网络相结合的残差时序卷积网络(Residual Temporal Convolutional Network, ResTCN)通信信号模式识别模型。该模型结构简单、网络收敛速度较快且具有较好的鲁棒性。通过实验仿真和海上试验对模型进行验证，在信噪比(Signal to Noise Ratio, SNR)大于-10 dB时，该方法在测试数据集上分类正确率为95%,在海试数据上正确率可达到93.5%。

Underwater communication signal pattern recognition plays an important role in non-cooperative underwater communication signal recognition. However, the complex and ever-changing underwater acoustic channel brings great challenges to underwater communication signal pattern recognition. To address the high model complexity and large number of extracted features of traditional algorithms, a Residual Temporal Convolutional Network(ResTCN)-based communication signal pattern recognition model is proposed, which combines residual network and time series convolutional network. The proposed model has a simple structure, high training convergence speed, and good robustness. The proposed model is verified in simulation and sea test experiments, and the result shows that the classification accuracy of the proposed method is 95% on the test data set and 93.5% on the sea test data, respectively, when the Signal to Noise Ratio(SNR) is greater than-10 dB.