An adaptive polarization correction algorithm for mitigating time-varying amplitude and phase errors

In dual-channel polarization diversity reception systems, amplitude and phase errors exhibit temporal variations due to factors such as mechanical antenna vibration, local oscillator drift, and gain fluctuations, leading to the distortion of polarization characteristics. In this paper, by constructing a time-varying amplitude and phase error model for dual-channel systems and analyzing the underlying error mechanisms, we design a joint blind equalization model in the polarization domain. A two-stage polarization adaptive filtering correction algorithm, termed TPAF, is proposed. The first stage involves variable-step-size polarization amplitude-phase momentum adaptive filtering (VPM-NLMS), which dynamically adjusts the iteration step size based on phase concentration and amplitude variance, while introducing a momentum mechanism to accelerate transient convergence, thereby achieving preliminary noise reduction and bias correction of the polarization sequence. The second stage employs a multi-constraint polarization correction autoencoder (PC-AE) for feature reconstruction. A polarization amplitude-phase correction loss function is proposed, which combines reconstruction loss, temporal smoothing loss, and phase concentration loss to stabilize the polarization fingerprint and separate transient noise using a convolutional encoder-decoder architecture, thereby enabling polarization feature reconstruction. Simulation results demonstrate that the proposed method can theoretically effectively suppress time-varying channel disturbances and significantly reduce estimation errors in polarization amplitude and phase. Real-world data results indicate that the algorithm presented in this paper exhibits a marked improvement over existing adaptive filtering algorithms, fully validating its reliability and practical value.