Acoustic emission technique can monitor the fatigue damage state of composite materials in real time. The identification accuracy of damage state depends on the processing and extraction of acoustic emission signals. A method of the acoustic emission signal processing based on harmonic wavelet power spectrum and energy spectrum is proposed. The damage state of composite materials with delamination defects can be identified. First, the uniaxial tensile experiment was carried out on the composite laminates with or without delamination defects. The acoustic emission signals during the process of damage evolution were recorded. Then, the frequency distribution range of different damage modes based on the harmonic wavelet packet decomposition was obtained according to the damage characteristics analyzed by frequency spectrum and energy spectrum. Compared with the traditional wavelet analysis, the harmonic wavelet analysis can detect the different damage characteristics and damage modes of composite blade with delamination defects accurately. It can be used to detect the operated state for in-service blades in practical engineering.