In long-span bridges and high-rise buildings, closely spaced modes are commonly observed, which greatly increases the challenge of identifying modal parameters. Hilbert–Huang transform (HHT), a widely used method for modal parameter identification, first applies empirical mode decomposition (EMD) to decompose the acquired response and then uses the Hilbert transform (HT) to identify the modal parameters. However, the problem is that the deficiency of mode separation of EMD in HHT limits its application for structures with closely spaced modes. In this study, an improved HHT based on analytical mode decomposition (AMD) is proposed and is used to identify the modal parameters of structures with closely spaced modes. In the improved HHT, AMD is first employed to replace EMD for decomposing the measured response into several mono-component modes. Then, the random decrement technique is applied to the decomposed mono-component modes to obtain the free decay responses. Furthermore, the resulting free decay responses are analyzed by HT to estimate the modal parameters of structures with closely spaced modes. Examples of a simple three-degree-of-freedom system with closely spaced modes, a high-rise building under ambient excitation, and the Ting Kau bridge under typhoon excitations are adopted to validate the accuracy, effectiveness, and applicability of the proposed method. The results demonstrate that the proposed method can efficiently and accurately identify the natural frequencies and damping ratios of structures with closely spaced modes. Moreover, its identification results are more precise compared to those obtained using existing methods.