To address the problem of response signal processing for high-altitude, long-endurance aircraft with high-aspect-ratio wings under turbulent excitation, this study proposes a modeling and quasi-modal parameter identification method that combines a Time-Varying Autoregressive (TVAR) model with the Unscented Kalman Filter (UKF). Within a state-space framework, the method dynamically estimates the TVAR coefficients and extracts quasi-modal frequencies and damping ratios from the structural responses, thereby effectively capturing the evolution of modal parameters under the combined effects of geometric nonlinearity and aerodynamic disturbances. To overcome the poor adaptability of traditional orthogonal basis function methods to local disturbances, the UKF is employed for recursive coefficient estimation, while probabilistic statistical features are introduced to globally characterize and assess the stability of the time-varying parameters. Wind tunnel experimental data validate that the proposed method exhibits high modeling accuracy and strong sensitivity to non-stationary responses, enabling accurate tracking of the dynamic evolution of each mode. Specifically, the first mode is characterized by a stable frequency and strong damping perturbations, dominating the flutter behavior; the second mode exhibits jumps and drifts, indicating a potential instability risk; while the third mode, despite having extremely low damping, does not participate in the instability process, reflecting a “high-frequency–low-coupling” feature typical of flexible structures. The results provide a high-resolution analytical tool for signal modeling and flutter identification of high-aspect-ratio aircraft under complex excitations, demonstrating strong engineering applicability and promising prospects for practical implementation.