LAUSR

The polarization attributes of seismic recordings are usually used to determine the seismic phase and to design a filter. However, different time window lengths lead to different time-domain polarization estimation results; the time window length is difficult to determine and has no uniform criterion. We propose a time-domain polarization estimation method employing an optimal time window based on the instantaneous frequency of the seismic signals. First, the time–frequency spectrum of the seismic signal is obtained using the continuous wavelet transform, and an optimal time window is constructed using the corresponding instantaneous dominant frequency. Then, a complex polarization analysis algorithm is used to estimate the polarization parameters. Compared with the conventional constant and center frequency time window, the polarization properties estimated using an optimal time window can achieve more accurate seismic phase identification. The picked arrival time error decreases from 82.4 to 32.0 s, about 60%. Furthermore, the corresponding polarization filter can effectively denoise noisy seismic recordings and increase the signal-to-noise ratio (SNR) increased by 0.49–2.65 times for original signals with different SNRs. The cross-correlation coefficient is increased by 0.26 on average, further preservation of signal information.