Modulation recognition method of mixed signals based on cyclic spectrum projection

Area of Science:

• Electrical Engineering
• Signal Processing
• Machine Learning

Background:

• Complex electromagnetic environments lead to mixed signals with various modulation types.
• Existing mixed signal recognition methods lack adaptability.
• Modulation recognition of mixed signals is a critical challenge in modern communications.

Purpose of the Study:

• To propose a novel and adaptable method for mixed signal modulation recognition.
• To leverage cyclic spectrum properties and deep neural networks for enhanced identification.
• To overcome limitations of existing methods in diverse signal conditions.

Main Methods:

• Theoretical derivation proving the feasibility of cyclic spectrum for signal identification.
• Utilizing grayscale projections of the 2D cyclic spectrum as identifying representations.
• Applying nonlinear piecewise mapping and directed pseudo-clustering to enhance spectral images.
• Employing deep neural networks for abstract feature extraction and recognition.

Main Results:

• The proposed method demonstrates robustness against noise, achieving over 95% average recognition rate at signal-to-noise ratios >= 0 dB.
• The technique shows good adaptability to variations in signal symbol rates.
• The method is resilient to changes in energy ratios between mixed signals.

Conclusions:

• The developed method effectively recognizes mixed signals using cyclic spectrum projection and deep neural networks.
• The approach offers significant improvements in adaptability and robustness compared to existing techniques.
• This method provides a reliable solution for modulation recognition in complex signal environments.