Abstract: Spectral decomposition decomposes a time domain signal to the time?frequency domain, which can characterize the local timefrequency properties and instantaneous frequency. The time?frequency resolution of conventional spectral decomposition methods is not high due to Heisenberg uncertainty principle. The reassigned spectrum method avoids the Heisenberg uncertainty principle and the timefrequency resolution is increased. Besides, it also generates concentrated energy for random noise. The multiple windows reassigned spectrum is based on reassigned spectrum and gets reassigned spectrums using orthogonality and recurrence. At last,the reassigned spectrums are distributed by the weighted value based on least square estimation. Synthetic data and field data show that the proposed approach can get high time?frequency resolution and disperse the random noise; meanwhile the real position of time?frequency energy can be located and indicates the low?frequency shadow related with hydrocarbon effectively