Terahertz-Band Channel and Beam Split Estimation via Array Perturbation Model

For the demonstration of ultra-wideband bandwidth and pencil-beamforming, terahertz (THz)-band has been envisioned as one key enabling technologies for sixth generation networks. However, acquisition THz channel entails several unique challenges such severe path loss beam-split. Prior works usually employ ultra-massive arrays additional hardware components comprised time-delayers to compensate these loses. In order provide a cost-effective solution, this paper introduces sparse-Bayesian-learning (SBL) technique joint beam-split estimation. Specifically, we first model an array perturbation inspired from signal processing. Next, low-complexity approach is developed by exploiting line-of-sight-dominant feature reduce computational complexity involved in proposed SBL estimation (SBCE). Additionally, based on federated-learning, implement model-free model-based SBCE solution. Further that, examine near-field considerations channel, introduce range-dependent The theoretical performance bounds, i.e., Cram\'er-Rao lower are derived both near- far-field parameters, e.g., user directions, ranges. Numerical simulations demonstrate that outperforms existing approaches exhibits cost.