Satellite Signal Modulation Pattern Recognition Based on Multi-feature Fusion

‍ ‍A modulation identification algorithm is proposed to address the problem of identifying QPSK-type modulation modes commonly used in satellite communications. The algorithm combines spectral， instantaneous statistical， higher-order cumulative， and amplitude distribution features. Based on signal pre-processing and frequency domain detection， the algorithm first uses rate signal and quadratic spectral features to identify the OQPSK and BPSK modulation modes. It then estimates parameters such as carrier frequency， bandwidth， and symbol rate using the spectral centroid and rate signal methods. It extracts higher-order cumulative feature parameters from the complex signal after orthogonal frequency down-conversion and matched filtering for symbol timing synchronization. The remaining signals are classified as either phase or amplitude shift keying signals. Finally， the algorithm uses fourth-order spectral and amplitude distribution features to identify all modulation modes. Simulation results showed that the algorithm is insensitive to frequency offsets and exhibits better identification performance for 16APSK and 32QAM than traditional methods based on higher-order spectra. Furthermore， at a signal-to-noise ratio of 6 dB， the algorithm achieved a 100% detection rate for modulation modes other than 16QAM and 64QAM. To validate the effectiveness of the algorithm， we constructed a test platform using a signal source and acquisition card. The classification results of nine actual acquisition signals confirmed the effectiveness of the algorithm. In addition， the presence of inherent errors in the carrier frequency estimation using the spectral centroid method highlighted the insensitivity of the algorithm to frequency offsets. In conclusion， the proposed algorithm combines multiple features and successfully identifies commonly used QPSK-type modulation modes in satellite communications. The algorithm exhibits robustness against frequency offsets and achieves high recognition rates for various modulation modes， making it a valuable solution for modulation identification in practical applications.