Signal fingerprint feature extraction and recognition method for communication satellite

Abstract

ABSTRACT Due to the complex maritime environment, communication satellite signals are inevitably subjected to various unpredictable situations which results in the noisy signal. To tackle the fickle ocean environment, we propose a novel scheme to extract and identify the fingerprint feature of communication satellite signals. K-means clustering algorithm is utilised to optimise the conventional Smooth Pseudo Wigner-Ville Distribution (SPWVD), which makes the final result of SPWVD closer to the feature of the original data and improves the accuracy of signal fingerprinting. Basing on the improved SPWVD, the singular value entropy is calculated by time-frequency analysis. Then, the entropy of diagonal slices is generated by the ameliorated bispectrum method which is enhanced by using the fractional low-order covariance method. Consequently, a two-dimensional signal fingerprint feature vector can be calculated. For the fingerprint feature recognition, the XGBoost classifier which is more efficient in machine learning classification is employed. Finally, we simulate the Iridium satellite signal for experiments with MATLAB. The experimental result shows the superiority of the improved SPWVD method and the improved bispectrum method, which proves the scheme's feasibility.