Time-frequency spectrum analysis can provide the signal features in time and frequency domains.However, the method is difficult to make a good tradeoff between high time-resolution and high frequency-resolution in the process of signal identification.Therefore, a new algorithm is proposed, which has a good time-frequency concentration.Firstly, the signal is decomposed into different frequency bands by harmonic wavelet packet.Then, Morlet wavelets with high time-resolution, and short-time Fourier transform with high frequency-resolution, are used to analyze the time-frequency distribution of each frequency band.Therefore, Morlet wavelets scale matrix and short-time Fourier time-frequency matrix are obtained.At last, these two matrices are fused into a new matrix with a specific algorithm.Analysis is made to accelerometer dynamic signal under simulation and dynamic condition.The result shows that the fusion algorithm can not only extract the weak dynamic change, but also has high time-frequency resolution, which is an intuitive, comprehensive, and accurate signal identification algorithm.