Abstract: A delay cluster selection optimization based second-order blind identification (DCSO-SOBI) carrier separation algorithm against distortion is proposed to address the demands of complex interference analysis and extraction in low Earth orbit (LEO) internet satellite systems. Targeting the limitations of traditional signal detection and recognition algorithms in handling multi-carrier aliasing scenarios, the proposed algorithm employes blind separation techniques to achieve the separation and extraction of multi-source mixed signals, thereby resolving the feature ambiguity issues caused by time-frequency aliasing of multiple signals. Furthermore, based on the correlation matrix characteristics of the observed signals within the second-order blind identification (SOBI) separation architecture, the proposed algorithm optimizes the initial selection of delay clusters and adjusts the search step size, thereby effectively reducing the search range and computational load of joint block diagonalization (JBD) while improving carrier separation accuracy and convergence speed. Simulation results demonstrate that, compared with the traditional SOBI algorithm, the proposed algorithm exhibits insensitivity to signal types，and achieves a 7.89% improvement in separation correlation coefficient and a 20.81% improvement in residual signal-to-noise ratio (SNR) under a 10dB SNR condition. In terms of computational complexity, the proposed algorithm achieves a significant reduction in the convergence speed of JBD processing at a relatively low computational cost for delay cluster selection optimization. Compared with the Jacobi-like JBD algorithm based on QR decomposition, the required number of iterations is reduced by 10.97%, and the computation time is improved by 0.62ms under a 10dB SNR condition, thereby effectively lowering the computational complexity and processing time required for separation. Without impacting the normal communication of LEO satellites, the proposed algorithm enables highprecision separation and rapid extraction of complex interference signals, providing a foundation for subsequent signal recognition processing and anti-interference scheme decision-making.

Key words: LEO internet satellite, carrier separation, anti-distortion, delay cluster selection optimization, second-order blind identification