This paper introduces a Wiener filtering method based on short time spectrum estimation which is implemented in frequency domain and applied to process seismic signals of high frequency. Since many seismic signal filtering methods could preserve the valid information of middle and low frequency well but could not preserve the valid information of high frequency well, so it needs to adopt a method which could preserve both low and high frequency of seismic signal. The Wiener filtering is the optimal filtering in the sense of minimum mean square error, and the way of frequency domain implementation is simpler and more effective than time domain implementation. So, it adopts Wiener filtering of frequency domain in this paper to obtain better results in seismic signal processing than conventional filtering method of time domain such as Kalman filtering, especially for high frequency seismic signals.