In this paper, based on the field data of wind wave in Tai Lake, power spectrum estimated by the Maximum Entropy Method(MEM) is compared with that estimated by using conventional methods. The results show that, the MEM is superior to the conventional methods of spectral estimation. The spectrum estimation by MEM avoids the shortcomings of using the window function and unrealistic assumptions, so that the spectral leakage is decreased and the accuracy of the spectral estimation is increased. It has better resolution and gives more realistic power estimates, especially for short data records. A new spectrum analytical technique for the time series data of wind wave is proposed.