A Blind-Manifold-Based Self-Calibration Method for Near-Field Uniform Linear Arrays With Sensor Position Errors

Abstract

Accurate near-field array measurement and sensing require precise knowledge of the array geometry, which is often degraded by sensor position perturbations. To address the strong parameter coupling in conventional joint-estimation approaches, this article proposes a blind-manifold-based self-calibration framework that decouples manifold reconstruction from geometric parameter estimation. An eigenvalue-based source-enumeration step is first incorporated for multisource operation. Blind signal separation then recovers the distorted array manifold, which is converted into geometry-consistent path-difference measurements. By exploiting the nonlinear distance-dependent phase structure of near-field propagation, the proposed method performs search-free calibration through linear least-squares initialization and nonlinear geometric refinement. From an instrumentation and measurement perspective, local identifiability, a path-difference-domain Fisher information formulation, and a first-order perturbation model are established to quantify achievable accuracy and error propagation. The validation is expanded to include cross-source tests, blind source separation (BSS) separability analysis, element-pattern phase-ripple stress tests, and measured-channel state information (CSI)-based robust geometry validation using a phase-stable OFDM subcarrier. The simulation results, semi-synthetic validation using measured waveforms, and complementary measured-CSI validation under real multiantenna phase residuals demonstrate improved geometry consistency, localization performance, and computational efficiency compared with conventional approaches.