Abstract: Advanced composite materials such as carbon fiber are widely employed in aerospace and other high-tech fields due to their superior properties; however, their low impact resistance makes quantitative evaluation of impact energy essential for ensuring structural safety and reducing maintenance costs. This paper proposes an impact-energy quantification method based on intelligent perception of spectral features. First, a physics-based model is established using multi-sensor spectral signatures to reliably characterize structural impact energy. An unsupervised adaptive cut-off frequency optimization strategy is then introduced, guided by the maximum coefficient of determination (R²), to achieve intelligent spectral feature perception. Finally, a constrained Iteratively Reweighted Least Squares (IRLS) algorithm is applied for robust model fitting, enabling accurate impact energy estimation. Experimental validation on T300 composite structures was performed with 14 energy levels (0.5~7 J) across multiple impact locations, assessed via leave-one-out cross-validation. Results demonstrate that the proposed method achieves average diagnostic accuracy above 90% for all impact positions, confirming its high precision and robustness.