利用标准时频变换方法在强噪声环境下无偏拾取地震P波、S波到时

地震P波、S波到时是精确分析地震水平位置、深度与速度结构等的重要参数，如何准确拾取P波和S波到时是地震学的一项重要的基础工作.大数据量与强噪声环境给地震到时的自动拾取带来了很大挑战.在频率域中可将信号与噪声分离，但会造成震相的偏移.针对上述问题，本文在STA/LTA、AIC方法的基础上，引入了标准时频变换（Normal Time-Frequency Transform，NTFT），结合信号时间域与频率域特征，提出了基于NTFT的STA/LTA方法，以及基于NTFT的AIC方法来拾取P波和S波的到时.基于NTFT的STA/LTA方法通过构建即时频率约束的特征函数，以增强地震信号振幅响应的变化特征.基于NTFT的AIC方法则根据NTFT的变换系数定位即时频率-时间基准点，通过滑动窗口直接对标准时频谱进行AIC处理拾取最佳到时.本文采用了不同强度噪声的60组合成数据和105组实测地震数据对方法的可靠性进行检验.以人工拾取到时为参考，实测数据中NTFT-STA/LTA方法拾取P波、S波到时的均方根误差分别为0.36 s和0.56 s；NTFT-AIC方法拾取P波、S波到时的均方根误差分别为0.25 s和0.35 s.相比于STA/LTA、AIC方法，NTFT改进后的方法提高了P波和S波到时的拾取准确率，为强噪声环境下的地震波形到时拾取提供了新思路.

Unbiased picking onset time of P and S phases by Normal Time-Frequency Transform method under a strong noise environment

Accurate arrival time of seismic phase is a significant parameter to accurately analyze the horizontal position, depth, and velocity structure of seismic. Therefore, how to accurately pick up the arrival times of P-wave and S-wave is the fundamental work in seismology. A large amount of data and strong noise bring great challenges to automatically picking up the arrival times of different phases. The frequency-domain can separate the signal from the noise, but it can cause the shift of the seismic phase. To solve the above problems, Normal Time-Frequency Transform (NTFT) is introduced based on Short-Term Average and Long-Term Average ratio (STA/LTA) and Akaike Information Criteria (AIC) methods. Our purpose is to combine the characteristics of the signal time domain with those of the frequency domain. Therefore, the NTFT-based STA/LTA method and NTFT-based AIC method are proposed to pick up the arrival times of P-wave and S-wave. The NTFT-based STA/LTA method enhances the amplitude response of seismic signals by constructing the characteristic function, which is constrained by the immediate frequency. In addition, the NTFT-based AIC method locates the instant frequency-time slider position according to the coefficients obtained from NTFT transformation firstly. Then, the sliding window is used to directly select the area of interest in the Normal Time-Frequency spectrum. Finally, the optimal arrival time of the seismic signal is picked up by the AIC method. The combined data of 60 and 105 sets of measured seismic data were analyzed. Comparing the NTFT-STA/LTA method with the manual method, the root-mean-square errors of P-wave and S-wave are 0.36 s and 0.56 s, respectively. Meanwhile, the root-mean-square errors of the P-wave and S-wave picked up by the NTFT-AIC method are 0.25 s and 0.35 s respectively. Compared with STA/LTA and AIC methods, two novel NTFT-based methods have higher accuracy, which provides a new solution to pick up different waveforms under a strong noise environment.