Although the zero crossing detection algorithm is being applied widely in frequency estimation of a single sinusoid for its low computation and for the convenience of realization，the zero crossing method is easily effected by the stochastic noise in the frequency estimation performance. In order to solve this problem, a new method is presented based on moving window data and zero crossing detection. Firstly, the multi-groups moving window data are obtained from the observation data; Secondly, the pre-estimated frequency values are obtained by using the zero crossing detection algorithm in multi-groups moving window data; Then, the outliers of frequency values are removed by threshold processing. Finally, the average value is confirmed as an estimated result. A series of frequency estimation factors are discussed in detail, such as data length, slipping times, adjoining sliding point number and sampling frequency. The result shows that the proposed method can improve the performance of frequency estimation. And the new method is feasible.