臺灣博碩士論文加值系統

在本論文中，我們利用類神經網路(ANN)模型來做一維跟二維的到達方向(DOA)估計，並且著重於低複雜度的模型。首先我們提出了兩種低複雜度的方法用來做一維到達方向估計，這兩種方法是分別利用已經存在的兩種低複雜度估計方法為基礎，並且各自加以改善而來。然後我們提出了一個低複雜度的方法用來做二維到達方向估計，關鍵在於我們把一個二維的到達方向估計問題轉換成兩個一維的到達方向估計問題。此外我們把轉換後，透過L型陣列得到的估計結果與透過X型陣列得到的估計結果做結合，以得到更好的表現。因此，我們可以用一維到達方向估計方法為基礎來實現二維到達方向估計，並且利用低複雜度的類神經網路模型。模擬結果顯示，在維持低複雜度模型的同時，我們提出的二維到達方向估計方法有更好的表現。

In the thesis, we consider direction-of-arrival (DOA) estimation with artificial neural networks (ANNs), for both one-dimensional (1D) and two-dimensional (2D), and focus on low-complexity models. First, we propose two methods for 1D multiple DOA estimation, improving the performance of existing low-complexity ANNs. Then, we proposed a low-complexity method for 2D multiple DOA estimation. The idea is to transform the 2D DOA estimation problem into two 1D DOA estimation. To further improve the performance, we combine the results from the L-shaped array and the X-shaped array. The proposed method is then implemented with an ANN. Simulations show that while the computational complexity remains low, the performance of the combined method can greatly enhance the performance of DOA estimation.

摘要 i
Abstract ii
Acknowledgement iii
Content iv
List of tables vi
List of figures vii
Chapter 1. Introduction 1
1.1 Motivation 1
1.2 Review 2
1.3 Proposed algorithms 4
1.4 Outline 5
Chapter 2. ULA DOA Estimation with Artificial Neural Networks 6
2.1 Multilayer perceptron neural network 6
2.1.1 Introduction to MLPNN 6
2.1.2 System modeling and data preprocessing 9
2.2 Sector-based method with rotation operation 11
2.2.1 System modeling and data preprocessing 11
2.2.2 Rotation operation 15
2.2.3 Spatial filter 17
2.2.4 Power re-normalization 20
2.2.5 Classifier 21
2.3 Sector-based method with DFT-nulling operation 24
2.3.1 System modeling and data preprocessing 24
2.3.2 Power detector 25
2.3.3 Classifier 27
2.3.4 DFT-nulling method 28
2.3.5 Signal shaping 29
Chapter 3. UPA DOA Estimation with Artificial Neural Networks 31
3.1 2D DOA estimation 31
3.1.1 Introduction to 2D estimation method 31
3.1.2 Introduction to the extended 1D method 33
3.2 2D estimation method with ANNs 33
3.2.1 System modeling and data preprocessing 33
3.2.2 2D sector-based method 38
3.3 Extended 1D method with L-shaped array 40
3.3.1 System modeling and data preprocessing 40
3.3.2 Angle-transformation 43
3.3.3 Pair-matching 44
3.4 Extended 1D method with X-shaped array 45
3.4.1 System modeling and data preprocessing 45
3.4.2 Angle-transformation 49
3.4.3 Pair-matching 51
3.5 Proposed 2D DOA estimation method with ANNs 52
3.5.1 Overlapping sector-based method with rotation operation 52
3.5.2 Sector-based method with DFT-partial nulling 54
3.5.3 System modeling and data preprocessing 55
3.5.4 Combined estimation with L-shaped and X-shaped array 55
Chapter 4. Simulations 58
4.1 1D DOA estimation 58
4.1.1 Simulation settings 58
4.1.2 MLPNN structures for SB-R 59
4.1.3 MLPNN structures for SB-DFT-N 61
4.1.4 Simulation results 62
4.2 2D DOA estimation 73
4.2.1 Simulation settings 73
4.2.2 Simulation results 74
Chapter 5. Conclusions 80
Reference 82

[1] R.O. Schmidt, "Multiple emitter location and signal parameter estimation," IEEE Trans. Antennas Propag., vol. 34, no. 3, pp. 276-280, March 1986.

[2] R. Roy and T. Kailath, "ESPRIT-estimation of signal parameters via rotational invariance techniques," IEEE Trans. Acoust. Speech Signal Process., vol. 37, no. 7, pp. 984-995, July 1989.

[3] H. L. Southall, J. A. Simmers and T. H. O'Donnell, "Direction finding in phased arrays with a neural network beamformer," IEEE Trans. Antennas Propag., vol. 43, no. 12, pp. 1369-1374, Dec. 1995.

[4] A. H. El Zooghby, C. G. Christodoulou and M. Georgiopoulos, "A neural network-based smart antenna for multiple source tracking," IEEE Trans. Antennas Propag., vol. 48, no. 5, pp. 768-776, May 2000.

[5] Z. Liu, C. Zhang and P. S. Yu, "Direction-of-Arrival Estimation Based on Deep Neural Networks With Robustness to Array Imperfections," IEEE Trans. Antennas Propag., vol. 66, no. 12, pp. 7315-7327, Dec. 2018.

[6] A. Faye, A. B. Youm and J. D. Ndaw, "LVQ Based DOA Estimation," Proc. 5th Int. Conf. Comput. Intell. Commun. Syst. Netw., Madrid, 2013, pp. 245-250.

[7] You-Yu Chen, "DOA Estimation with Artificial Neural Network," M. S. thesis, National Chiao Tung University, Taiwan, 2019.

[8] M. Agatonovic, Z. Stanković and B. Milovanović, "High resolution 2D DOA estimation using artificial neural networks," In Proc. European Conference on Antennas and Propagation, Prague, 2012, pp. 1-5.

[9] B. C. Csáji, "Approximation with artificial neural networks," M.Sc. thesis,
Faculty of Sciences, Eötvös Loránd University, Hungary, vol. 24, p. 48, 2001.

[10] Q. J. Zhang and K. C. Gupta, Neural networks for RF and microwave
design, Artech House, July 2000.

[11] C. G. Christodoulou and M. Georgiopoulos, Application of neural
networks in electromagnetics, Artech House, December 2000.


[12] J. Chen and H. Ma, "An accurate real-time algorithm for spectrum peaks search in 2D MUSIC," In Proc. International Conference on Multimedia Technology, Hangzhou, 2011, pp. 3385-3388.

[13] N. Tayem and H. M. Kwon, "L-shape 2-dimensional arrival angle estimation with propagator method," IEEE Trans. Antennas Propag., vol. 53, no. 5, pp. 1622-1630, May 2005.

[14] Y. Wei and X. Guo, "Pair-Matching Method by Signal Covariance Matrices for 2D-DOA Estimation," IEEE Antennas Wireless Propag. Lett., vol. 13, pp. 1199-1202, 2014.

[15] S. Kikuchi, H. Tsuji and A. Sano, "Pair-Matching Method for Estimating 2-D Angle of Arrival With a Cross-Correlation Matrix," IEEE Antennas Wireless Propag. Lett., vol. 5, pp. 35-40, 2006