Pilot-Assisted Hybrid Training Strategy With Orthogonal Polynomial Basis for Receiver Nonlinear Post-Distortion
Nonlinear distortions in receiver front-ends cause spectral regrowth and in-band distortion. Adaptive compensation is challenging because transmitted symbols are unknown at the receiver. This paper proposes a pilot-assisted hybrid training strategy that combines offline pre-training and online fine-tuning with a short pilot sequence using recursive least squares (RLS). To maximize parameter estimation efficiency, the post-distorter adopts a memory polynomial with an orthogonal basis. Orthogonality yields a well-conditioned input correlation matrix, dramatically accelerating RLS convergence and improving numerical stability. Simulations with 16-QAM OFDM signals demonstrate that the proposed method efficiently eliminates nonlinear distortions, significantly reducing pilot overhead and improving spectral purity compared to conventional memory polynomials. Its low complexity and online adaptability make it suitable for 5G/6G receivers.