Maritime Spectrum Sensing Using a Parallel CNN-LSTM Network

With the rapid development of wireless communication technology, the shortage of spectrum resources is becoming more and more prominent. Cognitive radio technology improves spectrum utilization by allowing secondary users to use idle frequency bands without interfering with primary users through spectrum sensing techniques. However, in complex marine environments, traditional spectrum sensing methods face challenges such as low signal-to-noise ratio, channel fading and noise interference, and detection performance is difficult to guarantee. In this paper, a marine spectrum sensing method based on parallel convolutional neural network-long short-term memory network (CNN-LSTM) is proposed. The method combines the spatial feature extraction capability of CNN and the time series modelling capability of LSTM, which can effectively learn the features of ocean signals and improve the accuracy and robustness of spectrum sensing. The experimental results show that the CNN-LSTM model has a higher detection probability in low SNR environments and is more sensitive to SNR, and can work more stably over a wider range of SNR than the traditional 1D CNN and LSTM models.