Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/74329Full metadata record
| ???org.dspace.app.webui.jsptag.ItemTag.dcfield??? | Value | Language |
|---|---|---|
| dc.contributor.advisor | 丁建均 | |
| dc.contributor.author | Chih-Hao Wang | en |
| dc.contributor.author | 王治皓 | zh_TW |
| dc.date.accessioned | 2021-06-17T08:30:07Z | - |
| dc.date.available | 2020-08-16 | |
| dc.date.copyright | 2019-08-16 | |
| dc.date.issued | 2019 | |
| dc.date.submitted | 2019-08-12 | |
| dc.identifier.citation | [1] D. Klein, “Lagrange multipliers without permanent scarring.” University of California at Berkeley, Computer Science Division (2004).
[2] Suykens, Johan AK, and Joos Vandewalle. ”Least squares support vector machine classifiers.”Neural processing letters 9.3 (1999): 293-300. [3] Tong, Simon, and Daphne Koller. ”Support vector machine active learning with applications to text classification.” The Journal of Machine Learning Research 2 (2002): 45-66. [4] http://www.cmlab.csie.ntu.edu.tw/~cyy/learning/tutorials/SVM1.pdf [5] Olaf Ronneberger, Philipp Fischer, and Thomas Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation.” University of Freiburg, Germany [6] https://towardsdatascience.com/types-of-convolutions-in-deep-learning-717013397f4d [7] B. R. Mahafza, Radar Systems Analysis and Design using MATLAB, Chapman and Hall, 2013 [8] P. Tait, Introduction to Radar Target Recognition, The Institution of Engineering and Technology, Jan 2006 [9] Ward, Keith D., Tough, Robert J. A., and Watts, Simon, Sea Clutter: Scattering, the K Distribution and Radar Performance, Inst of Engineering & Technology, June 2006 [10] Skolnik Merrill I. Introduction to Radar Systems, McGraw Hill; 1st Edition edition, 1962 [11] （美）陳，雷達中的微多普勒效應，電子工業出版社，2013年7月 [12] 翁文凱, 周宗仁, 尹彰, 邱永芳, & 何良勝. (2011). 利用海雜波推算推算海面波場特性. 海洋工程學刊, 11(1), 31-55. [13] Lingwei Ye, Dong Xia, Weibo Guo, “Comparison and Analysis of Radar Sea Clutter K Distribution Sequence Model Simulation Based on ZMNL and SIRP,” Modeling and Simulation, 7(01), 8, 2018. [14] 趙翠，海雜波特性及其抑制技術的研究，電子科技大學2014年碩士論文 [15] A. V. Oppenheim, Discrete-Time Signal Processing, 3rd Ed., Prentice Hall, New Jersey, 2010. [16] A. V. Oppenheim and A. S. Willsky, adapted by S. H. Nawab and J. J. Ding, Signals and Systems, 2nd ed. (adapted version), Eurosia Book Co., Taipei, Taiwan, Dec. 2016. [17] P. Zarchan, Tactical and Strategic Missile Guidance, AIAA, Washington, D. C., 1980. [18] R. E. Kalman, “A New Approach to Linear Filtering and Prediction Problems,” ASME Journal of Basic Engineering, Series 82d, pp.35-45, 1960. [19] S. Qian and D. Chen, Joint Time-Frequency Analysis: Methods and Applications, Prentice-Hall, 1996. [20] K. Grochenig, Foundations of Time-Frequency Analysis, Birkhauser, Boston, 2001. [21] L. Debnath, Wavelet Transforms and Time-Frequency Signal Analysis, Birkhäuser, Boston, 2001. [22] S. Mallat, A Wavelet Tour of Signal Processing: The Sparse Way, Academic Press, 3rd ed., 2009. [23] Jaeger, Herbert. Tutorial on Training Recurrent Neural Networks, Covering BPPT, RTRL, EKF and the Echo State Network Approach. GMD-Forschungszentrum Informationstechnik, 2002. [24] B. R. Mahafza, Radar Systems Analysis and Design Using MATLAB, Chapman and Hall/CRC, 2013. [25] V. C. Chen and H. Ling, Time-frequency Transforms for Radar Imaging and Signal Analysis, Artech House radar library, Artech House, Boston, MA, 2002. [26] Boggiatto, P., De Donno, G., & Oliaro, A., “Uncertainty principle, positivity and Lp-boundedness for generalized spectrograms,” Journal of Mathematical Analysis and Applications, vol. 335(1), pp. 93-112, 2007 [27] R. Altes, “Spectrograms and generalized spectrograms for classification of random processes,” In: Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP'84. IEEE, pp. 282-285, 1984. [28] Auger, François, and Patrick Flandrin. 'Improving the readability of time-frequency and time-scale representations by the reassignment method,' IEEE Transactions on Signal Processing, vol. 43, issue 5, pp. 1068-1089, 1995. [29] L. Teng, 'Dopplet performance analysis of drequency stepped radar signal,' Modern Radar, vol. 2, 1996. [30] D. A. Warde and S. M. Torres, “The autocorrelation spectral density for Doppler-weather-radar signal analysis,” IEEE Trans. Geoscience and Remote Sensing, vol. 52, issue 1, pp.508-518, Jan. 2014. [31] P. Suresh, T. Thayaparan, T, Obulesu, and K. Venkataramaniah, “Extracting micro-Doppler radar signatures from rotating targets using Fourier–Bessel transform and time–frequency analysis,” IEEE Trans. Geoscience and Remote Sensing, vol. 52, issue 6, pp. 3204-3210, 2014. [32] S. L. Marple, S. Barbarossa, B. G. Ferguson, K. W. Lo, G. J. Frazer, B. Boashash, V. Chandran, A. Gholami, and S. Ouelha, “Time-frequency methods in radar, sonar, and acoustics,” in Boualem Boashash (Ed.), Time-frequency Signal Analysis and Processing: A Comprehensive Reference, Second ed. (pp. 793-856) Amsterdam, Netherlands: Academic Press, 2016. [33] Kim, Youngwook, Sungjae Ha, and Jihoon Kwon. 'Human detection using Doppler radar based on physical characteristics of targets,' IEEE Geoscience and Remote Sensing Letters, vol. 12, issue 2, pp. 289-293, 2015 [34] B. R. Mahafza, Introduction to Radar Analysis, CRC press, 2017. [35] Sanghera, Jasbinder, Lynda Busse, and Ishwar Aggarwal, 'Missile warning and protection system for aircraft platforms,' U.S. Patent No. 6,873,893. 29 Mar. 2005. [36] Null, Fay E. 'Doppler countermeasure device,' U.S. Patent No. 4,149,166. 10 Apr. 1979. [37] Rougas, John A. 'Anti-radar missile (ARM) countermeasure method.' U.S. Patent No. 6,414,622. 2 Jul. 2002. [38] Gray, G. J., Aouf, N., Richardson, M. A., Butters, B., & Walmsley, R. H. (2013). Countermeasure effectiveness against an intelligent imaging infrared anti-ship missile. Optical Engineering, 52(2), 026401. [39] O'Neill, Mary Dominique. 'Method and apparatus for aircraft protection against missile threats.' U.S. Patent No. 6,410,897. 25 Jun. 2002. [40] Krupkin, Vladimir. 'Countermeasure system.' U.S. Patent No. 9,766,325, Sept. 2017. [41] Flandrin, Patrick, François Auger, and Eric Chassande-Mottin. 'Time-frequency reassignment: from principles to algorithms,' Applications in Time-Frequency Signal Processing, vol. 5, pp. 179-203, 2003. [42] Auger, F., Flandrin, P., Lin, Y. T., McLaughlin, S., Meignen, S., Oberlin, T., & Wu, H. T., “Time-frequency reassignment and synchrosqueezing: An overview,” IEEE Signal Processing Magazine, vol. 30, issue 6, pp. 32-41, 2013. [43] Lee, S. H., Hsiao, T. Y., & Lee, G. S., “Audio–vocal responses of vocal fundamental frequency and formant during sustained vowel vocalizations in different noises,” Hearing Research, vol. 324, pp. 1-6, 2015 [44] Gallagher, K. A., Narayanan, R. M., Mazzaro, G. J., Ranney, K. I., Martone, A. F., & Sherbondy, K. D., “Moving target indication with non-linear radar,” In Radar Conference (RadarCon), IEEE, pp. 1428-1433, May 2015. | |
| dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/74329 | - |
| dc.description.abstract | 我們主要的目標是偵測微型目標。在做微型目標的偵測時，首先，我們分析大約每隔 50 ~ 2000 ms 的間隔往各個不同方向發射的 narrow beam 的雷達信號，一個類似於方波或其他固定波形的信號。接著，我們再根據接收到的回傳的雷達信號的延遲，運用都卜勒效應 (Doppler effect) 來判斷物體的速度。然而，由於微型目標的體積小且距離遠，以致於它們的雷達截面積 (Radar Cross Section, RCS) 通常很小，不易做辨識。
而用來偵測微型目標的雷達信號，更是容易受到干擾。一般而言，雷達信號的訊噪比 (signal to noise ratio, SNR) 大約在 10dB 至 15dB 之間。這在訊號處理的問題當中，算是不高的，容易造成辨識的錯誤。 在此篇論文中，首先我們先使用時頻分析得到雷達信號的時頻圖。接著利用前處理的方式，如內插與異常資訊分析來改善我們的資料。藉由使用脊部濾波器(ridge filter)找出我們欲分析的信號，最後利用自相關與傅立葉轉換得到的頻譜來分析旋轉頻率。這個問題類似於找出音樂信號的基頻，會有倍頻的出現，我們設立四個條件來決定最終旋轉頻率。 另外我們也試著使用深度學習的方式(UNet)來分析雷達信號，由於訓練模型的需要，我們自行模擬資料產生許多輸入資料來幫助我們訓練模型，並對UNet架構進行修改使其更適合使用在雷達信號分析。同時我們也使用兩個經典的機器學習的方式，SVM與KNN來和UNet預測的結果做比較。我們希望使用深度學習的方式來減少雜訊對分析雷達信號的影響。 | zh_TW |
| dc.description.abstract | Our main purpose is to detect micro-target. In the detection of micro-targets, first, we analyze the radar signal of the narrow beam, which is transmitted at intervals of 50 to 2000 ms in different directions. The radar signal is similar to a square wave or other fixed waveform. Next, we use the Doppler effect to determine the velocity of the object based on the delay of the received radar signal. However, because micro-target is small and far away, its Radar Cross Section (RCS) is usually small and difficult to identify.
Radar signals used to detect micro-targets are more susceptible to interference. In general, the signal-to-noise ratio (SNR) of radar signals is between 10dB and 15dB. This is not high in the problem of signal processing, and it is easy to cause identification errors. In this thesis, first of all, we use the time-frequency analysis to obtain the time-frequency diagram of the radar signal. We use pre-processing methods such as interpolation and abnormal information analysis to improve our data. By using a ridge filter, we can find out the signal that we want to analyze. Finally, using the spectrum obtained by autocorrelation and Fourier transform to analyze the rotation frequency. This problem is similar to finding the fundamental frequency of a music signal, and there will be lots of harmonic frequencies. We set four conditions to determine the final rotation frequency. Besides, we also try to use the deep learning method (UNet) to analyze the radar signal. We generate a lot of input data to help us train the model and modify the UNet architecture to make it more suitable for radar signal analysis. At the same time, using two classic machine learning methods, SVM and KNN to compare with the results of the UNet prediction. We hope to use deep learning to reduce the impact of noise on analyzing radar signals. | en |
| dc.description.provenance | Made available in DSpace on 2021-06-17T08:30:07Z (GMT). No. of bitstreams: 1 ntu-108-R06942116-1.pdf: 3019012 bytes, checksum: db076519aefd00e206a603e45ed39a76 (MD5) Previous issue date: 2019 | en |
| dc.description.tableofcontents | CONTENTS
口試委員會審定書 A 誌謝 B 中文摘要 i ABSTRACT ii CONTENTS iv List of Figure vi List of Table xi Chapter 1 : Related Work 1 1.1 Support Vector Machine 1 1.2 the UNet 4 1.2.1 Introduction 4 1.2.2 Architecture 6 1.3 Time Frequency Analysis 7 Chapter 2 Overview of the Proposed Method 8 Chapter 3 Proposed Method 9 3.1 Time frequency analysis and radar signal pre-processing 9 3.2 Analysis of side spectral part amplitude and frequency 16 3.3 Simulation data 27 3.3.1 motivation 27 3.3.2 the method to make the simulation data 27 3.3.3 generate the ground truth 29 3.4 the Unet 30 3.4.1 the motivation for using the UNet 30 3.4.2 the modified Unet 31 3.5 SVM 34 3.5.1 ridge filter 34 3.5.2 smooth filter 36 3.6 KNN 37 Chapter 4 : Simulation result 38 Chapter 5 : Conclusion 59 REFERENCE 61 List of Figure Fig.1-1 To show the cutoff lines and the result of classification. 2 Fig.1-2 The figure of Support hyperplane 3 Fig. 1-3 (a)raw image (b) different color indicate different cells[5] 5 Fig. 1-4 The UNet architecture[5] 7 Fig. 1-5 The framework of the proposed method 9 Fig. 3-1 The time-frequency distribution of a normal data 11 Fig. 3-2 The time-frequency distribution of losing some of the data 12 Fig 3-3 The side spectral part of spectrum is too weak to be seen 12 Fig.3-4 After adjusting the energy between side spectral part and central spectral part. 14 Fig.3-5 After using the methods of step1 and step2 15 Fig.3-6 Some example of the time-frequency distribution of a normal data 15 Fig.3-7 Some example of the time-frequency distribution of a normal data 16 Fig.3-8 The space between two red lines is central spectral part 17 Fig.3-9 The projection value at the frequency axis from Fig.3-8 signal 18 Fig.3-10 The ridge region of Fig.3-8 20 Fig.3-11 The ridge region of the signal is projected along the frequency axis 20 Fig.3-12 The region between two red lines is central spectral part and the region between green line and red line is side spectral part 21 Fig.3-13 : (a) is Fig.3-1 signal projection value of the side spectral part ,(b) is the result of (a) to perform Fourier transform. 23 Fig.3-14 (a) Autocorrelation value of Fig.3-1 signal side spectral part, (b) is (a) to perform Fourier transform. 24 Fig.3-15 Some example of harmonic frequency. 25 Fig.3-16 The time frequency diagram for simulation data. 28 Fig.3-17 The time frequency diagram of data obtained by the simulation. 29 Fig.3-18 The ground truth of Fig.3-17 30 Fig.3-19 2D convolution with no padding, stride of 2 and kernel of 3[6] 32 Fig.3-20 Transposed 2D convolution with no padding, stride of 2 and kernel of 3[6] 32 Fig.3-21 The modified UNet architecture 33 Fig.3-22 The input simulation data with noise. 33 Fig.3-23 The left part of ground truth of Fig.3-22. 34 Fig.3-24 The result from the UNet. 34 Fig.3-25 The simulation data with noise 35 Fig.3-26 The result of performing the convolution along the vertical axis from ridge filter 35 Fig.3-27 The result of performing the convolution along the horizontal axis from ridge filter. 36 Fig.3-28 The result of performing the convolution along the vertical axis from smooth filter. 36 Fig.3-29 The result of performing the convolution along the horizontal axis from smooth filter. 37 Fig.3-30 The prediction of left side of the side spectral part from SVM model…37 Fig.3-31 The prediction of left side of the side spectral part from KNN model. 38 Fig.4-1 The time frequency diagram of real data 1 after using the post-processing. 39 Fig.4-2 The time frequency diagram of real data 2 after using the post-processing. 39 Fig.4-3 The time frequency diagram of real data 3 after using the post-processing. 40 Fig.4-4 The time frequency diagram of real data 4 after using the post-processing. 40 Fig.4-5 The time frequency diagram of real data 5 after using the post-processing. 41 Fig.4-6 The time frequency diagram of real data 6 after using the post-processing. 41 Fig.4-7 The time frequency diagram of real data 7 after using the post-processing. 42 Fig.4-8 The time frequency diagram of real data 8 after using the post-processing. 42 Fig.4-9 The time frequency diagram of real data 9 after using the post-processing. 43 Fig.4-10 The time frequency diagram of real data 10 after using the post-processing. 43 Fig.4-11 The time frequency diagram of real data 11 after using the post-processing. 44 Fig.4-12 The time frequency diagram of real data 12 after using the post-processing. 44 Fig.4-13 The time frequency diagram of real data 13 after using the post-processing. 45 Fig.4-14. The result of analyzing the rotation frequency for real data 1. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 45 Fig.4-15. The result of analyzing the rotation frequency for real data 2. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 46 Fig.4-16. The result of analyzing the rotation frequency for real data 3. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 46 Fig.4-17. The result of analyzing the rotation frequency for real data 4. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 47 Fig.4-18. The result of analyzing the rotation frequency for real data 5. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 47 Fig.4-19. The result of analyzing the rotation frequency for real data 6. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 48 Fig.4-20. The result of analyzing the rotation frequency for real data 7. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 48 Fig.4-21. The result of analyzing the rotation frequency for real data 8. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 49 Fig.4-22. The result of analyzing the rotation frequency for real data 9. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 49 Fig.4-23. The result of analyzing the rotation frequency for real data 10. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 50 Fig.4-24. The result of analyzing the rotation frequency for real data 11. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 50 Fig.4-25. The result of analyzing the rotation frequency for real data 12. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 51 Fig.4-26. The result of analyzing the rotation frequency for real data 13. (a) ridge detection (b) auto correlation (c) The Fourier transform of auto correlation(d)The result of the fundamental frequency detection after smoothing. 51 Fig.4-27 The time frequency diagram for simulation data 1. 52 Fig.4-28 The left one is the left part of side spectral part of Fig.4-27 of ground truth and the right one is the Unet prediction.. 53 Fig.4-29 The left one is the left part of side spectral part of Fig.4-27 of ground truth. the middle one is SVM prediction. The right one is KNN prediction. 53 Fig.4-30 The time frequency diagram for simulation data 2. 53 Fig.4-31 The left one is the left part of side spectral part of Fig.4-30 of ground truth and the right one is the Unet prediction. 53 Fig.4-32 The left one is the left part of side spectral part of Fig.4-30 of ground truth. The middle one is SVM prediction. The right one is KNN prediction. 54 Fig.4-33 The time frequency diagram for simulation data 3 54 Fig.4-34 The left one is the left part of side spectral part of Fig.4-33 of ground truth and the right one is the Unet prediction 54 Fig.4-35 The left one is the left part of side spectral part of Fig.4-33 of ground truth. The middle one is SVM prediction. The right one is KNN prediction 54 Fig.4-36 The time frequency diagram for simulation data 4 55 Fig.4-37 The left one is the left part of side spectral part of Fig.4-36 of ground truth and the right one is the Unet prediction 55 Fig.4-38 The left one is the left part of side spectral part of Fig.4-36 of ground truth. The middle one is SVM prediction. The right one is KNN prediction 55 Fig.4-39 The time frequency diagram for simulation data 4 56 Fig.4-40 The left one is the left part of side spectral part of Fig.4-39 of ground truth and the right one is the Unet prediction 56 Fig.4-41 The left one is the left part of side spectral part of Fig.4-39 of ground truth. The middle one is SVM prediction. The right one is KNN prediction 56 Fig.4-42 The time frequency diagram for simulation data 5 56 Fig.4-43 The left one is the left part of side spectral part of Fig.4-42 of ground truth and the right one is the Unet prediction 57 Fig.4-44 The left one is the left part of side spectral part of Fig.4-42 of ground truth and the right one is the Unet prediction 57 Fig.4-45 The time frequency diagram for simulation data 6 57 Fig.4-46 The left one is the left part of side spectral part of Fig.4-45 of ground truth and the right one is the Unet prediction 57 Fig.4-47 The left one is the left part of side spectral part of Fig.4-45 of ground truth and the right one is the Unet prediction 58 List of Table Table 1. the period of rotation (sec) 52 Table 2. the frequency of rotation 58 Table 3. the frequency of rotation 58 | |
| dc.language.iso | en | |
| dc.subject | 時頻分析 | zh_TW |
| dc.subject | 物件的偵測與識別 | zh_TW |
| dc.subject | 機器學習 | zh_TW |
| dc.subject | 異常資訊分析 | zh_TW |
| dc.subject | 背景雜訊處理 | zh_TW |
| dc.subject | 深度學習 | zh_TW |
| dc.subject | time-frequency analysis | en |
| dc.subject | machine learning algorithm | en |
| dc.subject | background noise processing | en |
| dc.subject | abnormal information analysis | en |
| dc.subject | object detection and recognition | en |
| dc.subject | deep learning algorithm | en |
| dc.title | 機器學習方法用於抗雜訊頻譜分析 | zh_TW |
| dc.title | Machine Learning Method for Noise Robust Frequency Analysis | en |
| dc.type | Thesis | |
| dc.date.schoolyear | 107-2 | |
| dc.description.degree | 碩士 | |
| dc.contributor.oralexamcommittee | 郭景明,許文良,歐陽良昱 | |
| dc.subject.keyword | 深度學習,背景雜訊處理,異常資訊分析,機器學習,物件的偵測與識別,時頻分析, | zh_TW |
| dc.subject.keyword | deep learning algorithm,machine learning algorithm,background noise processing,abnormal information analysis,object detection and recognition,time-frequency analysis, | en |
| dc.relation.page | 64 | |
| dc.identifier.doi | 10.6342/NTU201902987 | |
| dc.rights.note | 有償授權 | |
| dc.date.accepted | 2019-08-12 | |
| dc.contributor.author-college | 電機資訊學院 | zh_TW |
| dc.contributor.author-dept | 電信工程學研究所 | zh_TW |
| Appears in Collections: | 電信工程學研究所 | |
Files in This Item:
| File | Size | Format | |
|---|---|---|---|
| ntu-108-1.pdf Restricted Access | 2.95 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.