Varying-Window-Length Time-Frequency Peak Filtering and its Application to Seismic Data

The varying-window-length time-frequency peak filtering (VWL-TFPF) is proposed in this paper to promote its ability of removing random noise and holding higher frequency component of non-linear signal. TFPF has good performance of removing random white noise and enhancing band-limited deterministic signal, but it maybe distort the waveform of complex nonlinear signal with various dominant frequencies, such as seismic signal, when it remove the random noise. PWVD should be used to ensure the linear IF condition of TFP, thus the windowing length is important to remove random noise and to preserve the waveform of signal. In this paper, the relationship between window-length, sampled frequency and dominant frequency of signal is analyzed. According to the relationship, VWL-TFPF method chose different time-window-length at different time section of signal to resolve the distortion of high frequency components resulted from unsuitable window length of TFPF. This method don¿t need design the frequency band of filter and little prior condition of signal. Testing of this method is made on synthetic data and common shot point seismic data. The results show that VWL-TFPF technique has better performance on attenuating random noise and holding frequency character of filtered seismic reflected signal than TFPF.