Monitoring structural dynamic responses is an important factor in structural health monitoring. When a building is subjected to certain vibration excitations, the structure will produce certain vibration responses. Once a structure is damaged, variations occur in its natural frequencies, stiffness, and damping values. The accurate determination of structural damage is a difficult engineering problem in the monitoring of structural dynamic responses. In this paper, we propose a method for extracting structural damage information and identifying structural damage based on the wavelet packet energy eigenvector. Utilizing the decomposed energy eigenvalue of each frequency band as a structural dynamic parameter, we determined differences in the structural state by the observed variation in the energy value and obtained good results. We also analyzed the experimental data with respect to the influence of noise on energy values. In an experiment, we found that the energy eigenvector can be used to identify the structural state and was free of noise. However, it could not be used to extract the energy feature vector after de-noising the observation signal. Therefore, we propose a method for accurately determining situations with structural damage, which is practical and can facilitate structural damage monitoring.