臺灣博碩士論文加值系統

近年來因為電腦科技進步，神經網路的各種應用也再次蓬勃發展，很多研究者已經將神經網路的技術擴展到各自的領域，當然無線通訊領域也不例外，雖然已經有應用在無線通訊系統的上層(Upper layers)，但實體層(Physical layer)的應用會因為複雜通道環境的阻礙而有所限制，實作起來相對困難，儘管如此還是相信神經網路能夠提出有用且有見解的解決方案，並且有望能在難以用數學模型描述的通訊場景中有所突破。而本論文針對兩個實體層上和通道相關的模組，分別為通道估測與秩指標及預編碼矩陣指標選擇，嘗試使用神經網路的解決方案來處理。
本論文的第一個主題是基於神經網路的通道估測，將通道頻率響應視為一張二維影像，利用影像上超解析度(Super-resolution)的技術，藉由一個統一的神經網路(Unified neural network)，優化傳統使用內插得到的通道，並可以獲得更平滑的通道頻率響應，且降低通道估測誤差以及提升位元錯誤率(bit error rate)品質。最後發現在長延遲擴展造成嚴重的頻率選擇性衰減的通道能有顯著的效能增加，但是所需要的複雜度比起傳統通道估測卻是較高的，而這在卷積神經網路的解決方案當中是一個很難避免的問題。
本論文的第二個主題是基於自組織特徵映射圖的秩指標及預編碼指標選擇，因為需要對多個預編碼矩陣計算複數矩陣乘法與矩陣反矩陣或是矩陣行列式，所以傳統搜尋的複雜度非常大，而隨著天線數的增加或是天線的擺放方式不同，預編碼矩陣的碼簿大小也會急遽增加，所以本論文提出一低複雜度的解決方案，使用和以往完全不同的通道共變異數分群做法，對不同的多輸入多輸出相關性通道分群，並建立秩指標與預編碼矩陣指標的查找表，來完成秩指標與預編碼矩陣指標選擇，且在可容忍範圍內的效能降低，來達到降低運算複雜度的目的。

Recently, due to the advancement of computer technology, various applications of neural networks have flourished again. Researchers from different fields have extended neural networks to their respective fields, and of course, wireless communication is no exception. Although neural networks have been applied to the upper layer of the wireless communication system, applying them to the physical layer is quite challenging due to the sophistication of channel environments, which renders implementation tougher. Nevertheless, we still believe that neural networks can provide useful and insightful solutions, and they are expected to make a breakthrough in the communication scenarios that can hardly be expressed by mathematical models. In this thesis, we focused on the two channel-related modules of the physical layer processing, which are channel estimation and RI/PMI selection, respectively, and tried to process them with neural network solutions.
The first topic in this thesis is neural network-based channel estimation. We regard the channel frequency response(CFR) as a 2D image and utilize the super-resolution technique, which was originally used on images, and optimize the CFR obtained by traditional interpolation methods with a unified neural network to obtain a smoother CFR. This technique helps reduce channel estimation error and improves bit error rate quality. In the end, it was found that decoding performance for channels with highly frequency selective fading caused by long delay spread has improved significantly, but the required complexity is higher than traditional channel estimation approaches. However, this is usually an inevitable issue for CNN type solutions.
The second topic in this thesis is SOFM-based RI/PMI selection. The traditional RI/PMI selection approach utilizes all precoding matrices to calculate the complex matrix multiplications and matrix inversion or matrix determinant, which will lead to huge complexity. As the number of antennas increases or the arrangement of antennas varies, the size of the precoding matrix codebook will increase drastically. Therefore, we proposed a low complexity solution, which is the channel covariance matrix clustering. This approach is completely different from past approaches in that it groups different MIMO correlation channels and builds RI/PMI look-up tables for RI/PMI selection. In summary, this approach is a low-complexity solution with very similar performance for RI/PMI selection.

致謝 i
摘要 iii
Abstract v
目錄 vii
圖目錄 xii
表目錄 xvii
第一章 緒論 19
1.1 研究背景 19
1.2 研究動機 20
1.3 論文架構 22
第二章 5G-NR及LTE之相關標準介紹 23
2.1 3GPP標準簡介 24
2.1.1 訊框架構(Frame Structure) 24
2.1.2 波型參數(Numerology) 28
2.1.3 參考訊號(Reference Signal) 31
2.2 3GPP 多路徑衰減通道(Multipath Fading Channel)模型簡介 34
2.2.1 E-UTRA Channel models 34
2.2.2 NR Tapped Delay Line (TDL) models 35
2.2.3 MIMO Correlation Matrices 36
2.3 層映射與預編碼簡介 39
2.3.1 層映射(Layer mapping) 40
2.3.1.1 Layer mapping for transmission on a single antenna port 40
2.3.1.2 Layer mapping for spatial multiplexing 40
2.3.2 預編碼(Precoding) 41
2.3.2.1 Precoding for transmission on a single antenna port 41
2.3.2.2 Precoding for spatial multiplexing 41
2.3.3 預編碼碼簿(Codebook)介紹 42
2.3.3.1 LTE Codebook 42
2.3.3.2 NR Type I Single-Panel Codebook 44
第三章 神經網路介紹 47
3.1 多層感知器 (Multilayer Perceptron, MLP ) 47
3.1.1 架構 47
3.1.2 訓練與推理 50
3.2 深度卷積神經網路 (Deep Convolution Neural Network, DCNN ) 52
3.2.1 原理 52
3.2.2 架構 53
3.2.2.1 卷積層(Convolution layer) 53
3.2.2.2 池化層(Pooling layer) 56
3.2.3 訓練與推理 56
3.3 自組織特徵映射圖(Self-Organizing Feature Map, SOFM) 60
3.3.1 原理 60
3.3.2 架構 61
3.3.3 訓練與推理 62
3.3.3.1 初始化權重(initialize weights) 63
3.3.3.2 尋找最佳匹配神經元(Search for BMN) 63
3.3.3.3 決定最佳匹配神經元鄰近的神經元(Determine the BMN neighborhood) 64
3.3.3.4 更新權重(Update weights) 65
3.3.4 樹狀結構自組織特徵映射圖(Tree-Structured SOFM) 66
第四章 訓練與測試資料集之建立 69
4.1 傳送接收機系統架構 69
4.1.1 符元邊界粗估 (Coarse Symbol Boundary Detection, CSBD) 70
4.1.2 分數載波頻率飄移估測 (Fractional CFO Estimation) 72
4.1.3 快速傅立葉轉換(Fast Fourier Transform, FFT) 73
4.1.4 傳統通道估測(Channel Estimation, CE) 73
4.1.4.1 最小平方法(LS)通道估測 76
4.1.4.2 最小均方誤差法(MMSE)通道估測 77
4.1.4.3 內插(Interpolation) 79
4.1.5 等化(Equalization) 81
4.1.5.1 強制歸零(Zero Forcing, ZF)等化器 81
4.1.5.2 最小均方誤差法(MMSE)等化器 82
4.2 多路徑衰減通道資料集(Multipath Fading Channel Dataset) 82
4.2.1 訓練資料、測試資料與標準答案(Training data, Testing Data, and Golden data) 83
4.3 多輸入多輸出相關性資料集(MIMO Correlation Dataset) 87
4.3.1 訓練資料與測試資料(Training data and Testing Data) 87
第五章 基於神經網路之通道估測設計 91
5.1 超解析度(Super-Resolution) 91
5.2 本論文提出的通道估測 92
5.2.1 使用資料集 95
5.2.2 使用神經網路架構 96
5.2.2.1 卷積神經網路架構(CNN Architecture) 97
5.2.2.2 殘差學習架構(Residual Learning Architecture) 97
5.2.2.3 U-Net Architecture 99
5.3 模擬結果與結論 101
5.3.1 LTE Reference Signal Pattern 102
第六章 基於自組織特徵映射圖之秩指標及預編碼矩陣指標選擇設計 107
6.1 秩指標及預編碼矩陣指標選擇準則 107
6.2 傳統碼簿搜尋 110
6.2.1 窮盡搜尋(Exhaustive Search) 110
6.2.2 循序搜尋(Sequential Search) 111
6.3 本論文提出的秩指標及預編碼矩陣指標選擇 111
6.3.1 使用資料集 113
6.3.2 通道共變異數矩陣分群(Channel Covariance Matrix Clustering) 114
6.3.3 建立查找表 (Build Look-Up-Table) 116
6.3.4 樹狀搜尋(Tree Search) 116
6.3.4.1 餘弦相似度(Cosine Similarity) 117
6.3.4.2 通道容量(Channel Capacity) 117
6.4 模擬結果和複雜度分析與結論 118
6.4.1 MIMO channel without fading 119
6.4.2 MIMO channel with fading 123
6.4.2.1 平均12個subcarriers的做法 133
6.4.3 運算複雜度與查找表大小分析 137
6.4.3.1 運算複雜度分析 137
6.4.3.2 查找表大小分析 140
第七章 結論與展望 143
參考文獻 147

[1]Jeff Dean. 2017. AIFrontiers: Trends and Developments in DeepLearning Research [Online]. Available: https://www.slideshare.net/AIFrontiers/jeff-dean-trends-and-developments-in-deep-learning-research
[2]T. Wang, C. Wen, H. Wang, F. Gao, T. Jiang, and S. Jin, “Deep learning for wireless physical layer: Opportunities and challenges,” China Communications, vol. 14, no. 11, pp. 92-111, Nov. 2017.
[3]T. O’Shea and J. Hoydis, “An introduction to deep learning for the physical layer,” IEEE Transactions on Cognitive Communications and Networking, vol. 3, no. 4, pp. 563-575, Dec. 2017.
[4]E. Nachmani, Y. Be''ery, and D. Burshtein, “Learning to decode linear codes using deep learning,” in Proc. of 54th Annual Allerton Conference on Communication, Control, and Computing (Allerton), Monticello, IL, USA, Sept. 2016, pp. 341-346.
[5]E. Nachmani, E. Marciano, D. Burshtein, and Y. Be’ery, “RNN decoding of linear block codes,” arXiv preprint arXiv:1702.07560, 2017.
[6]T. Gruber, S. Cammerer, J. Hoydis, and S. t. Brink, “On deep learning-based channel decoding,” in Proc. of 51st Annual Conference on Information Sciences and Systems (CISS), Baltimore, MD, USA, March 2017, pp. 1-6.
[7]H. Ye, G. Y. Li, and B. Juang, “Power of Deep Learning for Channel Estimation and Signal Detection in OFDM Systems,” IEEE Wireless Communications Letters, vol. 7, no. 1, pp. 114-117, Feb. 2018.
[8]N. Samuel, T. Diskin, and A. Wiesel, “Deep MIMO detection,” in Proc. of IEEE 18th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), Sapporo, Japan, July 2017, pp. 1-5.
[9]T. J. O’Shea, T. Erpek, and T. C. Clancy, “Deep learning based MIMO communications,” arXiv preprint arXiv:1707.07980, 2017.
[10]S. Dörner, S. Cammerer, J. Hoydis, and S. t. Brink, “Deep Learning Based Communication Over the Air,” IEEE Journal of Selected Topics in Signal Processing, vol. 12, no. 1, pp. 132-143, Feb. 2018.
[11]A. Felix, S. Cammerer, S. Dörner, J. Hoydis, and S. Ten Brink, “OFDM-Autoencoder for End-to-End Learning of Communications Systems,” in Proc. of IEEE 19th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), Kalamata, Greece, June 2018, pp. 1-5.
[12]C. Dong, C. C. Loy, K. He, and X. Tang, “Image Super-Resolution Using Deep Convolutional Networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 38, no. 2, pp. 295-307, Feb. 2016.
[13]4G vs 5G [Online]. Available: https://www.sharetechnote.com/html/5G/5G_4Gvs5G.html
[14]W. C. Hu, Design and Implementation of Interference-Aware Receiver in Ultra Dense Network, M.S. thesis, National Taiwan University, Taipei, Taiwan, December 2017.
[15]5G NR Resource Grid [Online]. Available: http://www.sharetechnote.com/html/5G/5G_ResourceGrid.html
[16]S. Sesia, I. Toufik, and M. Baker, Eds., LTE: The UMTS Long Term Evolution. John Wiley and Sons, 2009.
[17]NR; Physical channels and modulation, 3GPP, TS 38.211 v15.7.0, Oct. 2019 [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/38211.htm
[18]5G NR Frame Structure [Online]. Available: http://www.sharetechnote.com/html/5G/5G_FrameStructure.html
[19]Downlink Frame Structure [Online]. Available: https://www.sharetechnote.com/html/FrameStructure_DL.html
[20]N. H. Huang, User Activity Detection and Pilot Sequence Design for Uplink Grant-free NOMA in 5G Networks, M.S. thesis, National Taiwan University, Taipei, Taiwan, August 2018.
[21]Evolved Universal Terrestrial Radio Access (E-UTRA); User Equipment (UE) radio transmission and reception, 3GPP, TS 36.101 v16.3.0 Sep. 2019 [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/36101.htm
[22]Study on channel model for frequencies from 0.5 to 100 GHz, 3GPP, TR 38.901 v16.0.0 Oct. 2019 [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/38901.htm
[23]Study on scenarios and requirements for next generation access technologies, 3GPP, TR 38.913 v15.0.0 July 2018 [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/38913.htm
[24]R1-143883, "High-level views on FD-MIMO and elevation beamforming", 3GPP RAN1 #78bis, Samsung, Oct. 2014
[25]Evolved Universal Terrestrial Radio Access (E-UTRA); Physical channels and modulation, 3GPP, TS 36.211 v15.7.0 Sep. 2019 [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/36211.htm
[26]Evolved Universal Terrestrial Radio Access (E-UTRA); Physical layer procedures, 3GPP, TS 36.213 v15.7.0 Sep. 2019 [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/36213.htm
[27]NR; Physical layer procedures for data, 3GPP, TS 38.214 v15.7.0 Sep. 2019 [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/38214.htm
[28]Sigmoid [Online]. Available: https://sebastianraschka.com/faq/docs/logisticregr-neuralnet.html
[29]V. Nair and G. E. Hinton, “Rectified linear units improve restricted boltzmann machines,” in Proc. of 27th International Conference on Machine Learning, 2010, pp. 807-814.
[30]ReLU : Not a Differentiable Function: Why used in Gradient Based Optimization? and Other Generalizations of ReLU.
[31]Batch gradient descent [Online]. Available: https://www.coursera.org/learn/machine-learning/lecture/9zJUs/mini-batch-gradient-descent
[32]Stochastic gradient descent [Online]. Available: http://ufldl.stanford.edu/tutorial/supervised/OptimizationStochasticGradientDescent/
[33]T. Kohonen, “The self-organizing map,” Proceedings of the IEEE, vol. 78, no. 9, pp. 1464-1480, Sept. 1990.
[34]SOFM Wiki [Online]. Available: https://en.wikipedia.org/wiki/Self-organizing_map
[35]Kohonen''s Self Organizing Feature Maps [Online]. Available:
http://www.ai-junkie.com/ann/som/som1.html
[36]Y. H. Sun and T. D. Chiueh, “A Flexible and High-Performance Self-Organizing Feature Map Training Acceleration Circuit and Its Applications,” in Proc. of IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS), Hsinchu, Taiwan, March 2019, pp. 92-96.
[37]T. D. Chiueh, T. T. Tang, and L.G. Chen, “Vector quantization using tree-structured self-organizing feature maps,” IEEE Journal on Selected Areas in Communications, vol. 12, no. 9, pp. 1594-1599, Dec. 1994.
[38]T. D. Chiueh, P. Y. Tsai, and I. W. Lai, Baseband Receiver Design for Wireless MIMO-OFDM Communications. Singapore: Wiley, Apr. 2012.
[39]C. W. Chu, C. C. Lee, and Y. H. Huang, “Design of an OFDMA Baseband Receiver for 3GPP-LTE Evolution,” in Proc. of IEEE International Symposium on VLSI Design, Automation and Test (VLSI-DAT), Hsinchu, Taiwan, Apr. 2008, pp. 169-199.
[40]C. C. Lee, C. F. Liao, C. M. Chen, and Y. H. Huang, “Design of 4x4 MIMO-OFDMA Receiver with Precode Codebook Search for 3GPP-LTE,” in Proc. of IEEE International Symposium on Circuits and Systems (ISCAS), Paris, France, May 2010, pp. 3957 - 3960.
[41]Channel Estimation [Online]. Available: http://www.sharetechnote.com/html/Communication_ChannelEstimation.html
[42]Y. S. Cho, J. Kim, W. Y. Yang, and C. G. Kang, MIMO-OFDM Wireless Communications with MATLAB. Singapore: Wiley (Asia) Pte. Ltd., 2010.
[43]J. Rinne and M. Renfors, “Pilot spacing in orthogonal frequency division multiplexing systems on practical channels,” IEEE Transactions on Consumer Electronics, vol. 42, no. 4, pp. 959-962, Nov. 1996.
[44]L. Erup, F. M. Gardner, and R. A. Harris, “Interpolation in digital modems. II. Implementation and performance,” IEEE Transactions on Communications, vol. 41, no. 6, pp. 998-1008, June 1993.
[45]W. Yang, X. Zhang, Y. Tian, W. Wang, J. Xue, and Q. Liao, “Deep Learning for Single Image Super-Resolution: A Brief Review,” IEEE Transactions on Multimedia, vol. 21, no. 12, pp. 3106-3121, Dec. 2019.
[46]K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. of the IEEE conference on computer vision and pattern recognition (CVPR), 2016, pp. 770-778.
[47]S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” arXiv preprint arXiv:1502.03167, 2015.
[48]J. Kim, J. K. Lee, and K. M. Lee, “Deeply-recursive convolutional network for image super-resolution,” in Proc. of the IEEE conference on computer vision and pattern recognition (CVPR), 2016, pp.1637-1645.
[49]G. E. Hinton and R. R. Salakhutdinov, “Reducing the dimensionality of data with neural networks,” Science, vol. 313, no. 5786, pp. 504–507, Jul. 2006.
[50]O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Proc. of Int. Conf. Medical Image Comput. Computer-Assisted Intervention, 2015, pp. 234-241.
[51]F. Penna, H. Cheng, and J. Lee, “Low Complexity Precoder Selection for FD-MIMO Systems,” in Proc. of IEEE 84th Vehicular Technology Conference (VTC-Fall), Montreal, QC, Canada, Sept. 2016, pp. 1-5.
[52]R1-062650, “Codebook Design for E-UTRA MIMO Pre-coding”, 3GPP RAN1 #46bis, Texas Instruments, Oct, 2006.
[53]Y. Lin, Y. Chen, C. Chu, C. Zhan, and A. Wu, “Dual-Mode Low-Complexity Codebook Searching Algorithm and VLSI Architecture for LTE/LTE-Advanced Systems,” IEEE Transactions on Signal Processing, vol. 61, no. 14, pp. 3545-3562, July 2013.
[54]Z. Bai et al., “Closed loop transmission with precoding selection in LTE/LTE-Advanced system,” in Proc. of 3rd International Symposium on Applied Sciences in Biomedical and Communication Technologies (ISABEL 2010), Rome, Italy, Nov. 2010, pp. 1-5.
[55]A. Wagdy, T. Khattab, and E. A. Sourour, “Modified QR-D and MMSE PMI selection technique for MIMO closed loop spatial multiplexing in LTE/LTE-advanced,” in Proc. of 7th IEEE GCC Conference and Exhibition (GCC), Doha, Qatar, Nov. 2013, pp. 93-97.
[56]S. Schwarz, M. Wrulich, and M. Rupp, ‘‘Mutual information based calculation of the precoding matrix indicator for 3GPP UMTS/LTE,’’ in Proc. of Int. ITG Workshop Smart Antennas (WSA), Bremen, Germany, Feb. 2010, pp. 52–58.
[57]DeSieno, “Adding a conscience to competitive learning,” in Proc. of IEEE 1988 International Conference on Neural Networks, San Diego, CA, USA, July 1988, pp. 117-124.
[58]Matrix determinant complexity [Online]. Available: https://en.wikipedia.org/wiki/Computational_complexity_of_mathematical_operations