Physics-embeded Multicarrier Signal Recognition for Ka-band Land Mobile Satellite Communications Systems under Rain Attenuation and Shadowed-Rice Fading

In Ka-band land mobile satellite (LMS) communication systems, accurate multicarrier signal recognition (MSR) is crucial for detecting malicious radio signal interference. However, prior research has predominantly focused on terrestrial wireless multicarrier systems, often neglecting the substantial impact of weather-related factors such as rain attenuation on MSR algorithms. To bridge this gap, this paper proposes a Multi-Focus Denoised Cyclic Autocorrelation-Liquid Time Constant Network (MFDCA-LTCNet) algorithm for MSR, specifically designed to mitigate the effects of rain attenuation and Ricean shadowing fading. We first develop a cyclic autocorrelation (CA) denoising algorithm to effectively reduce noise interference in multicarrier signal features. Additionally, a DCT-based multi-focus denoised CA feature fusion algorithm is proposed, integrating Principal Component Analysis (PCA), wavelet transform, and contourlet transform to counteract residual channel fading effects from multiple perspectives. Next, our MFDCA-LTCNet incorporates a physics-embedded liquid dynamic learning mechanism, leveraging first-order ordinary differential equations to achieve a lightweight and robust MSR classifier. Monte Carlo simulations demonstrate the effectiveness and superiority of the proposed MFDCA-LTCNet over existing MSR methods in mitigating adverse Shadowed-Rice (SR) channel effects and rainfall attenuation for Ka-band LMS systems.