Modulation Recognition of Underwater Acoustic Signals Using Deep Hybrid Neural Networks

It is a huge challenge for the receiver to correctly identify the modulation types due to the complex underwater channel environment and severe noise interference, while real-time communications have strict time requirements. To address this issue, this work combines the automatic feature extraction and learning ability of recurrent neural network (RNN) and convolutional neural network (CNN) to design a modulation recognition model for underwater acoustic signals, named recurrent and convolutional neural network (R&CNN). Compared with traditional modulation recognition techniques, this method achieves higher recognition accuracy without manual feature extraction. The experimental results show that the validation accuracy of the proposed R&CNN reaches 98.21% on the Trestle dataset and 99.38% on the South China Sea dataset, with an average recognition time of 7.164 ms. Compared with conventional deep learning methods, the proposed R&CNN not only has higher recognition accuracy, but also greatly reduces the recognition time.

IEEE Transactions on Wireless Communications