臺灣在地理上坐落於因板塊運動而造成地震事件頻繁的環太平洋火山帶上，由於擁有大量的地震發生，分析並了解板塊的運動顯得格外重要。通常來說密度越高的地震測站監測網絡可以更詳細地提供感測到的信號，而且地震的震波到達時間通常是由地質領域的專家依據多年的經驗以及該領域的知識所判斷出來的。隨著測站的變多，產生的波形圖也隨之變多，這使得人工標記的速度越來越難跟上波形圖增加的速度，並且不同專家之間的標準也會有所不同。因此有許多方法被開發以用於自動標記地震波的到達時間，然而傳統方法通常需要針對不同測站裡不同的傳感器設置各自的參數，並且每隔一段時間就要對參數進行調整以保持準確性，這也使得要維護持續增加的儀器成為一件負擔重大的工作。本文提出了一個基於深度卷積神經網路的深度學習方法，該方法不再需要為每個不同的儀器調整參數，並且同時能保持高準確率和低時間誤差。此方法利用的模型使用了臺灣的中央氣象局於2016年到2018年之間所記錄的地震圖來訓練和評估。在實驗中，此方法在標記地震到時上擁有低時間誤差，同時在判斷雜訊上擁有高準確度，而且此方法在與傳統自動標記方法比較下擁有更好的震波到時召回率並且不需要針對不同測站調整參數。

Taiwan is geographically located on the ``Ring of Fire'', where earthquakes occur frequently due to the movement and collision between plates. With a large number of earthquake events, it is important to analyze the process to understand the movement of the plates. In general, a monitoring network with high seismogram density is useful to provide the sensed signals in more detail. The arrival time is usually picked by personnel based on their years of experience and domain knowledge. As the number of waveforms produced by sensors continues to increase, it is difficult for domain experts to keep up with the ever-increasing data, and the standards may vary. Therefore, many methods have developed to pick the phase arrival time automatically, however, traditional auto picking methods usually require parameters individually for each sensor at each station, and the parameters need to be tuned every period of time to keep the accuracy, which will make maintaining the method in the expansion of the observation stations a huge effort. In this thesis, a deep learning method based on deep convolutional neural network (DCNN) is presented that no longer needs to adjust parameters for every different sensors, and remains high accuracy with low error time residual. It was trained and evaluated using the seismograms recorded by the Central Weather Bureau (CWB) in Taiwan from 2016 to 2018. In the experiments, the method has a low time residual in picking arrival time and a high accuracy in distinguishing noise. Moreover, the proposed method is compared with the traditional picking method and performs a better phase arrival time recall rate and there is no need to tune the parameters of each stations.

[1] J. S. Gomberg, K. M. Shedlock, and S. W. Roecker, “The effect of S-wave arrival times on the accuracy of hypocenter estimation,” Bulletin of the Seismological Society of America, vol. 80, no. 6A, pp. 1605–
1628, Dec. 1990.
[2] R. V. Allen, “Automatic earthquake recognition and timing from single traces,” Bulletin of the Seismological Society of America, vol. 68, no. 5, pp. 1521–1532, Oct. 1978.
[3] D.-Y. Chen, N.-C. Hsiao, and Y.-M. Wu, “The earthworm based earthquake alarm reporting system in Taiwan,” Bulletin of the Seismological Society of America, vol. 105, no. 2A, Mar. 2015.
[4] R. Sleeman and T. van Eck, “Robust automatic p-phase picking: an on-line implementation in the analysis of broadband seismogram recordings,” Physics of the Earth and Planetary Interiors, vol. 113, no. 1-4, pp. 265 – 275, Jun. 1999.
[5] N. Huang, Z. Shen, S. Long, M. Wu, H. Shih, Q. Zheng, N.-C. Yen, C.-C. Tung, and H. Liu, “The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis,” Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, vol. 454, no. 1971, pp. 903–995, Mar. 1998.
[6] R. Zhang and L. Zhang, “Method for identifying micro-seismic P-arrival by time-frequency analysis using intrinsic time-scale decomposition,” Acta Geophysica, vol. 63, no. 2, pp. 468–485, Apr. 2015.
[7] I. Kirbas and M. Peker, “Signal detection based on empirical mode decomposition and teager–kaiser energy operator and its application to P and S wave arrival time detection in seismic signal analysis,” Neural Computing and Applications, vol. 28, no. 10, pp. 3035–3045, Oct. 2017.
[8] M. Liu, J. Yang, Y. Cao, W. Fu, and Y. Cao, “A new method for arrival time determination of impact signal based on HHT and AIC,” Mechanical Systems and Signal Processing, vol. 86 (Part A), pp. 177 – 187, Mar. 2017.
[9] X. Li, X. Shang, A. Morales-Esteban, and Z. Wang, “Identifying P phase arrival of weak events: The akaike information criterion picking application based on the empirical mode decomposition,”Computers & Geosciences, vol. 100, pp. 57 – 66, Mar. 2017.
[10] H. Zhang, C. Thurber, and C. Rowe, “Automatic P-wave arrival detection and picking with multiscale wavelet analysis for single-component recordings,” Bulletin of the Seismological Society of America, vol. 93, no. 5, pp. 1904–1912, Oct. 2003.
[11] J. J. Galiana-Merino, J. L. Rosa-Herranz, and S. Parolai, “Seismic P phase picking using a kurtosis-based criterion in the stationary wavelet domain,” IEEE Transactions on Geoscience and Remote Sensing, vol. 46, no. 11, pp. 3815–3826, Nov. 2008.
[12] C. D. Saragiotis, L. J. Hadjileontiadis, and S. M. Panas, “A higher-order statistics-based phase identification of three-component seismograms in a redundant wavelet transform domain,” in Proceedings of the IEEE Signal Processing Workshop on Higher-Order Statistics, Jun. 1999, pp. 396–399.
[13] C. D. Saragiotis, L. J. Hadjileontiadis, and S. M. Panas, “Pai-s/k: A robust automatic seismic P phase arrival identification scheme,” IEEE Transactions on Geoscience and Remote Sensing, vol. 40, no. 6, pp. 1395–1404, Jun. 2002.
[14] C. Baillard, W. C. Crawford, V. Ballu, C. Hibert, and A. Mangeney, “An automatic kurtosis-based P- and S-phase picker designed for local seismic networks,” Bulletin of the Seismological Society of America, vol. 104, no. 1, pp. 394–409, Nov. 2013.
[15] J. VanDecar and R. S. Crosson, “Determination of teleseismic relative phase arrival times using multichannel cross-correlation and least squares,” Bulletin of the Seismological Society of America, vol. 80, no. 1, pp. 150–169, Feb. 1990.
[16] S. J. Gibbons and F. Ringdal, “The detection of low magnitude seismic events using array-based waveform correlation,” Geophysical Journal International, vol. 165, no. 1, pp. 149–166, Apr. 2006.
[17] E. H. Ait Laasri, E.-S. Akhouayri, D. Agliz, and A. Atmani, “Automatic detection and picking of P-wave arrival in locally stationary noise using cross-correlation,” Digital Signal Processing, vol. 26, pp. 87 – 100, Mar. 2014.
[18] I. Álvarez, L. García, S. Mota, G. Corté, C. Benítez, and Á. De la Torre, “An automatic P-phase picking algorithm based on adaptive multiband processing,” IEEE Geoscience and Remote Sensing Letters, vol. 10, no. 6, pp. 1488–1492, Nov. 2013.
[19] D. Shen, Q. Zhang, T. Xu, H. Zhu, W. Zhao, Z. Yin, P. Zhou, L. Fang, E. Chen, and H. Xiong, “A machine learning-enhanced robust P-phase picker for real-time seismic monitoring,” ArXiv e-prints, 2020.
[20] Z. E. Ross, M.-A. Meier, and E. Hauksson, “P wave arrival picking and first-motion polarity determination with deep learning,” Journal of Geophysical Research: Solid Earth, vol. 123, no. 6, pp. 5120–5129, Jun. 2018.
[21] W. Zhu and G. C. Beroza, “Phasenet: a deep-neural-network-based seismic arrival-time picking method,” Geophysical Journal International, vol. 216, no. 1, pp. 261–273, Oct. 2018.
[22] L.-C. Chen, Y. Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder-decoder with atrous separable convolution for semantic image segmentation.” in The European Conference on Computer Vision, Sep. 2018.
[23] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in IEEE Conference on Computer Vision and Pattern Recognition, Jun. 2016, pp. 770–778.
[24] F. Yu and V. Koltun, “Multi-scale context aggregation by dilated convolutions,” in International Conference on Learning Representations, May 2016.
[25] S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in Proceedings of the 32nd International Conference on Machine Learning, vol. 37, Jul. 2015, pp. 448–456.
[26] V. Nair and G. E. Hinton, “Rectified linear units improve restricted boltzmann machines,” in International Conference on Machine Learning, Jun. 2010.
[27] D. Bahdanau, K. Cho, and Y. Bengio, “Neural machine translation by jointly learning to align and translate” in International Conference on Learning Representations, May 2015.
[28] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. u. Kaiser, and I. Polosukhin, “Attention is all you need,” in Advances in Neural Information Processing Systems, vol. 30, Dec. 2017, pp. 5998–6008.
[29] J. S. Bridle, “Probabilistic interpretation of feedforward classification network outputs, with relationships to statistical pattern recognition,” in Neurocomputing, 1990, pp. 227–236.
[30] D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” in International Conference on Learning Representations, May 2015.
[31] D. E. Rumelhart, G. E. Hinton, and R. J. Williams, “Learning representations by back-propagating errors,” Nature, vol. 323, no. 6088, pp. 533–536, Oct. 1986.
[32] A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. Kopf, E. Yang, Z. DeVito, M. Raison, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala, “Pytorch: An imperative style, high-performance deep learning library,” in Advances in Neural Information Processing Systems, vol. 32, Dec. 2019, pp. 8026–8037.