Maritime Spectrum Sensing Based on Cyclostationary Features and Convolutional Neural Networks.

Area of Science:

• Electrical Engineering
• Computer Science
• Signal Processing

Background:

• Maritime cognitive radio networks (MCRN) face significant spectrum sensing (SS) challenges due to environmental dynamics.
• Existing SS methods struggle with the complexities of maritime environments, including sea movement and unstable links.

Purpose of the Study:

• To develop an adaptive spectrum sensing algorithm for MCRN that addresses environmental complexities.
• To improve the accuracy and reliability of spectrum sensing in maritime settings.

Main Methods:

• A classification-guided approach transforming SS into a classification problem.
• Extraction of cyclostationary features using Fast Fourier Transform (FFT) Accumulation Method (FAM).
• Utilization of Convolutional Neural Networks (CNNs) for signal classification and detection (TC2NND algorithm).

Main Results:

• The TC2NND algorithm achieved a 91.5% detection probability with a 5% false-alarm probability at -10 dB SNR.
• Demonstrated superior performance compared to conventional spectrum sensing methods in maritime environments.

Conclusions:

• The proposed TC2NND algorithm offers a robust and effective solution for spectrum sensing in MCRN.
• Leveraging cyclostationary features and CNNs significantly enhances SS performance in dynamic maritime conditions.