This paper adopts Hilbert-Huang Transform and Random Decrement Technique to evaluate amplitude-dependent damping of a steel building from strong motion records. The results demonstrate that amplitude-dependent damping is an inherent nature of steel buildings. Damping first increases with amplitude at the initial stage, then starts decreasing with amplitude after reaching a "critical amplitude". Based on Hilbert-Huang Transform and Random Decrement Technique, it is feasible to identify nonlinear damping ratios with respect to amplitude from strong motion records.