Generative Diffusion-Based Channel Estimation for Pinching Antenna-Assisted Indoor NFC

Pinching antenna systems (PASS) are waveguide-based antenna architectures featuring structural flexibility and high energy efficiency, making them attractive for indoor near-field communication (NFC). However, rich multipath propagation and spatial non-stationarity in practical indoor environments pose significant challenges to accurate channel estimation, especially under limited antenna activation and pilot resources. In this paper, the PASS channel estimation problem is reformulated from a generative inference perspective. A diffusion-model-driven channel estimation framework is proposed, where received signals are interpreted as noisy observations of latent near-field channel states, and channel estimation is performed via conditional reverse denoising diffusion. By exploiting waveguide-mediated near-field structures and sparse antenna activation, the proposed framework enables robust channel recovery in highly underdetermined settings. To better match indoor propagation characteristics, the diffusion-based inference emphasizes multipath-aware channel distributions, allowing joint modeling of deterministic waveguide effects and stochastic scattering, thereby alleviating model mismatch in conventional estimators. Simulation results show that the proposed method achieves stable channel estimation performance across different SNRs and antenna activation scales, while the computational complexity of the proposed framework is explicitly analyzed to assess its practical applicability.