本論文提出了兩種深度學習(DL)方法來區分不同的抽頭延遲線(TDL)通道模型，然後獲得相應長度的內部保護間隔(GI)，從而使離散傅利葉轉換傳播正交分頻多工(DFT-s-OFDM)系統可以獲得更好的頻譜效率性能並保持低錯誤率 (BER)。
頻譜效率(SE)是第五代新空中介面(5G NR)通信的重要發展指標。在傳統的多用戶正交分頻多工(OFDM)上行鏈路系統中，固定長度的循環前綴(CP)方法會導致頻譜效率丟失。提高頻譜效率的一種方法是調整保護間隔的長度。
在典型的正交分頻多工系統模型中，為了解決同步和符號間干擾(ISI)問題，循環前綴由於其循環卷積以及額外序列的特性而被利用。然而，多用戶上行系統中的每個用戶設備(UE)都應該設置相同大小的循環前綴，這會導致頻譜效率的嚴重浪費。因此，我們選擇無循環前綴的離散傅利葉轉換傳播正交分頻多工系統，用內部保護間隔代替循環前綴。與循環前綴相比，內部保護間隔的長度可以隨每個用戶設備變化以顯著提高頻譜效率。另外，深度學習在數據量充足的情況下具有很高的預測準確性和學習能力。它在許多識別和分類任務中都有出色的表現。因此將深度學習方法引入 5G 無線通信是一種吸引人的趨勢。
在本論文中，我們研究了幾種深度學習神經網路演算法來設計所提出的框架。通過訓練估計的通道頻域信息特徵，兩者都可以及時識別通道模型，從而有效地給出合適的內部保護間隔長度。數值結果表明，所提出的深度學習分類器能夠實現對多路徑通道模型進行分類以提供相應長度的保護間隔的高精確度。而且，可以在不嚴重影響錯誤率性能的情況下提高頻譜效率的表現。

This thesis proposes two deep learning (DL) approaches to distinguish the different tapped delay line (TDL) channel model then obtain the corresponding length of internal guard interval (GI), whereby the DFT-s-OFDM system can get better SE performance and maintain the low bit error rate (BER).
Spectral efficiency (SE) is a significant development indicator in 5G New Radio (NR) communication. In a conventional multi-user OFDM uplink system, the fixed length cyclic prefix (CP) approach results in the loss of SE. One way to enhance the SE is adapting the length of guard interval.
In typical OFDM system model, to deal with the problem of synchronization and inter-symbol interference (ISI), the CP is utilized due to its property of circular convolution and extra sequence. However, every user equipment (UE) in a multi-user uplink system should set the same size of CP which causes substantial waste of SE. Hence, we present the CP free DFT-s-OFDM system, which replaces CP with internal GI. Compared to CP approach, internal GI length can adjust adaptively with each UE to improve SE performance substantially. Additionally, deep learning has great prediction accuracy and learning ability with sufficient amount of data. It has excellent performance in many recognition and classification tasks. For this reason, leading in the deep learning method to 5G wireless communication is an attractive trend.
In this thesis, we study several DL neural network algorithms to design proposed framework. By means of training the estimated channel frequency-domain information feature, both of them can identify the channel model promptly to give the appropriate internal GI length effectively. The numerical results show that the proposed DL classifiers are able to achieve high accuracy about classifying the multipath channel models to supply the corresponding length of GI. Furthermore, the performance of SE can be improved without influencing BER performance seriously.

Chinese Abstract i
English Abstract Contents ii
Contents iv
1 INTRODUCTION 1

1.1 Foreword . . . . . . . . . . . .1
1.2 Research Motivation and Objective . . . . . . . 2
1.3 Related Work . . . . . . . . . . . . . . . . . 3
1.3.1 CP Based OFDM System . . . . . . . . . 3
1.3.2 CP Free OFDM System . . . . . . . . 3
1.4 Proposed Method . . . . . . . . . . . . . 4
1.5 Contribution and Achievement . . . . . . . . 5
1.6 Thesis Organization . . . . . . . . . . . . 6
2 BACKGROUNDS 7


2.1 Cyclic Prefix Discrete Fourier Transform Spread Orthogonal Frequency Divi- sion Multiplexing (CP DFT-s-OFDM) 7
2.2 Generalized Unique Word Discrete Fourier Transform Spread Orthogonal Fre- quency Division Multiplexing (GUW DFT-s-OFDM) 11
2.3 Multi-user DFT-s-OFDM Uplink 14
2.4 Time Dispersion Parameters Estimation 16

3 DEEP LEARNING BASED CHANNEL CLASSIFIER DESIGN FOR MULTI-USER CP FREE DFT-s-OFDM UPLINK SYSTEM 18

3.1 Transceiver System Model . . . . . . . . . . . . 18
3.2 Multipath Channel Establishment . . . . . . . . . 20
3.3 DL Based Channel Classifier Design with Off-line Training . 22
3.3.1 Training Dataset . . . . . . . . . . . . . . . 22
3.3.2 The Proposed Deep CNN Architecture . . . . . . . 23
3.3.3 The Proposed Deep RNN Architecture . . . . . . . 26
3.4 Model Training Setting . . . . . . . . . . . . . 31
3.5 Adaptive Internal GI Design . . . . . . . . . . . 32

4 NUMERICAL AND SIMULATION RESULTS 35

4.1 Simulation Parameters . . . . . . . 35
4.2 Confusion Matrix and F1-score . . . . . 36
4.3 Comparison . . . . . . . . . . . . . . . 38
4.4 Bit Error Rate Performance . . . . . . . 39


4.5 Spectral Efficiency 40
5 CONCLUSIONS 43
Bibliography 44

[1]M. Series, “Imt vision–framework and overall objectives of the future development of imt for 2020 and beyond,” Recommendation ITU, vol. 2083, p. 0, 2015.
[2]G. L. Stuber, J. R. Barry, S. W. Mclaughlin, Y. Li, M. A. Ingram, and T. G. Pratt, “Broadband mimo-ofdm wireless communications,” Proceedings of the IEEE, vol. 92, no. 2, pp. 271–294, 2004.
[3]H. G. Myung, J. Lim, and D. J. Goodman, “Single carrier fdma for uplink wireless transmission,” IEEE vehicular technology magazine, vol. 1, no. 3, pp. 30–38, 2006.
[4]B. Farhang-Boroujeny, “Ofdm versus filter bank multicarrier,” IEEE signal processing magazine, vol. 28, no. 3, pp. 92–112, 2011.
[5]J. Abdoli, M. Jia, and J. Ma, “Filtered ofdm: A new waveform for future wireless systems,” in 2015 IEEE 16th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), pp. 66–70, IEEE, 2015.
[6]V. Vakilian, T. Wild, F. Schaich, S. ten Brink, and J.-F. Frigon, “Universal-filtered multi-carrier technique for wireless systems beyond lte,” in 2013 IEEE Globecom Work- shops (GC Wkshps), pp. 223–228, IEEE, 2013.
[7]T. Wild, F. Schaich, and Y. Chen, “5g air interface design based on universal filtered (uf-
) ofdm,” in 2014 19th International Conference on Digital Signal Processing, pp. 699– 704, IEEE, 2014.
[8]K. Mizutani and H. Harada, “Universal time-domain windowed ofdm,” in 2016 IEEE 84th Vehicular Technology Conference (VTC-Fall), pp. 1–5, IEEE, 2016.
[9]G. Berardinelli, K. I. Pedersen, T. B. Sorensen, and P. Mogensen, “Generalized dft- spread-ofdm as 5g waveform,” IEEE Communications Magazine, vol. 54, no. 11, pp. 99– 105, 2016.
[10]U. Kumar, C. Ibars, A. Bhorkar, and H. Jung, “A waveform for 5g: Guard interval dft-s-ofdm,” in 2015 IEEE Globecom Workshops (GC Wkshps), pp. 1–6, IEEE, 2015.
[11]R. Odarchenko, O. Tkalich, Z. Shevchuk, and S. Lukin, “Ofdm signal formation with adaptive guard interval duration change,” in 2015 Second international scientific- practical conference problems of infocommunications science and technology (PIC S&T), pp. 216–219, IEEE, 2015.
[12]I. Goodfellow, Y. Bengio, and A. Courville, Deep learning. MIT press, 2016.
[13]S.-S. Wei and J.-M. Wu, “Multiple objective optimization of osbe and isi for cyclic prefix free dft-s-ofdm systems,” in 2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring), pp. 1–5, IEEE, 2020.
[14]H. Yin and S. Alamouti, “Ofdma: A broadband wireless access technology,” in 2006 IEEE sarnoff symposium, pp. 1–4, IEEE, 2006.
[15]G. Berardinelli, L. A. M. R. de Temino, S. Frattasi, M. I. Rahman, and P. Mogensen, “Ofdma vs. sc-fdma: Performance comparison in local area imt-a scenarios,” IEEE Wireless Communications, vol. 15, no. 5, pp. 64–72, 2008.
[16]G. Berardinelli, F. M. Tavares, T. B. Sørensen, P. Mogensen, and K. Pajukoski, “Zero-tail dft-spread-ofdm signals,” in 2013 IEEE Globecom Workshops (GC Wkshps), pp. 229–234, IEEE, 2013.
[17]M. Huemer, C. Hofbauer, A. Onic, and J. B. Huber, “Design and analysis of uw-ofdm signals,” AEU-International Journal of Electronics and Communications, vol. 68, no. 10, pp. 958–968, 2014.
[18]A. Sahin, R. Yang, M. Ghosh, and R. L. Olesen, “An improved unique word dft-spread ofdm scheme for 5g systems,” in 2015 IEEE Globecom Workshops (GC Wkshps), pp. 1– 6, IEEE, 2015.
[19]T. Yucek and H. Arslan, “Time dispersion and delay spread estimation for adaptive ofdm systems,” IEEE transactions on vehicular technology, vol. 57, no. 3, pp. 1715– 1722, 2008.
[20]C. Tepedelenlioglu and G. B. Giannakis, “On velocity estimation and correlation proper- ties of narrow-band mobile communication channels,” IEEE Transactions on Vehicular Technology, vol. 50, no. 4, pp. 1039–1052, 2001.
[21]Y. S. Cho, J. Kim, W. Y. Yang, and C. G. Kang, MIMO-OFDM wireless communica- tions with MATLAB. John Wiley & Sons, 2010.
[22]G. T. 38.901, “Study on channel model for frequencies from 0.5 to 100 ghz,” 2017.
[23]D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014.