監督式學習應用於多正交分頻多工系統之通道延遲擴展分類

簡易檢索 / 詳目顯示

回結果列表

研究生：	加尼 Jami, Hema Ganesh
論文名稱：	監督式學習應用於多正交分頻多工系統之通道延遲擴展分類 Supervised Learning Classification of Channel Delay Spread for Variable Guard Interval to OFDM Systems
指導教授：	吳仁銘 Wu, Jen-Ming
口試委員:	伍紹勳 WU, SAU HSUAN 鍾偉和 Chung, Wei-Ho
學位類別：	碩士 Master
系所名稱：	電機資訊學院 - 通訊工程研究所 Communications Engineering
論文出版年：	2020
畢業學年度：	109
語文別：	英文
論文頁數：	45
中文關鍵詞：	於正交分頻多工、可調式保護間距、用戶的平均過量延、k-近鄰演算法
外文關鍵詞：	Variable Guard Interval, KNN,SVM
相關次數：	點閱：2 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

這篇論文主題的目標是表現在基於正交分頻多工(OFDM)的系統，使用可調式保護間距(variable guard interval)的多項優點。迴圈字首(CP)對於CP-OFDM跟DFT-s-OFDM是外加的固定保護間距，而迴圈字首會降低系統的頻譜效益(SE)。為了克服這個缺點，ZT-DFT-s-OFDM、UW-DFT-s-OFDM以及GUW-DFT-s-OFDM被提出，而這些方法是使用內部的保護間距(IGI)，而內部的保護間距可以改善外加的保護間距(CP)上述所產生的問題，進而改善頻譜效益。然而內部的保護間距沒有迴圈字首所擁有的圓周摺積(circular convolution)特性，追蹤通道的延遲擴展(delay spread)對於系統很困難因此造成額外的頻譜浪費。為了解決這樣的問題，提出”在正交分頻多工系統下通道延遲擴展的監督式學習分類方法用於可調式保護間隔”。
雖然每個用戶的平均過量延遲(mean excess delay)，但是每個用戶的延遲概觀(delay profile)。因此即使先前的模型使用彈性的保護間隔，在不考慮功率延遲概觀的情況下，使用簡單的參數，例如平均延遲，來決定保護間隔是不充分的。這讓我們有動機使用機器學習的技術來決定可調式保護間隔。
一開始，我們基於第三代合作夥伴計畫(3GPP)協定裡所定義的延遲線模型(TDL model)的正規畫延遲以及接收功率，藉此產生隨機的功率延遲概觀資料。藉由使用資料編排技術，每個用戶的功率延遲概觀資料被簡化為方均根延遲擴展(RMS delay spread)，以減少計算上的複雜度。接著使用機器學習的技術來訓練資料，進而做不同類別的分類，最後用這個分類後的結果來決定每個用戶的可調式保護間隔。我們使用的架構是GUW DFT-s-OFDM，將內部的保護間隔換成這裡提出的可調式保護間隔，以此來檢查所提出的模型的表現。
最後分類的結果顯示出，K-近鄰演算法(KNN)做出的預測正確率會優於使用支撐向量機演算法(SVM)的結果。而且模擬結果指出我們所提出的方法有較好的位元錯誤率(BER)以及頻譜效益。

The objective of this thesis is to feature the advantages of utilizing the Variable Guard
Interval (VGI) to OFDM based systems. A cyclic prefix (CP) which was fixed is used
as the external guard interval for CP-OFDM and CP-Discrete Fourier Transform spread
Orthogonal Frequency Division Multiplexing (DFT-s-OFDM) subsequently decreases the
Spectral Efficiency (SE) of the systems. To overcome this, the Internal Guard Interval
(IGI) is developed for Zero Tail (ZT)-DFT-s-OFDM, Unique word (UW)-DFT-s-OFDM and
Generalized unique word (GUW)-DFT-s-OFDM are proposed to overcome the problem of
CP. It provides the same functionality of providing a guard period as CP. This IGI improves
the SE of the systems by stagnating the total symbol duration. However, IGI does not provide
exact circular convolution compared with CP due to the non-cyclic leakage. Tracking the
delay spread of the channel is difficult for the system causes the extra overhead. To overcome
this problem ”Supervised learning classification of channel delay spread for variable guard
interval to OFDM systems” is proposed. Using this approach benefits the model with better
flexibility than previous guard periods.
Though the mean excess delay is the same for every user, the delay profile will be
different from user to user. So, even previous models use a flexible guard interval, they
are sophisticated because it is insufficient to use simplified parameters like mean delay to
determine the guard interval without considering power delay profile. This motivates us to
use the learning-based technique to determine the variable guard interval.
i
So the contribution to this thesis is explained as, we first generated the random data of
power delay profile based on Normalized delay and received Power for Tapped Delay Line
(TDL) models from A to E using 3GPP specifications. Using the data formatting technique,
the power delay profile is simplified to the RMS delay spread for every user of TDL models
to reduce the computational complexity for the classification. As the data is supervised, the
Machine learning classification technique is used for training the data to classify the different
classes which can later be used to determine the variable guard interval (VGI) to the users
based on RMS delay spread. We consider G-UW DFT-s-OFDM as the application to replace
the internal guard interval with a variable guard interval to check the performance of our
proposed model.
The classification results show that the k-nearest neighbors (KNN) algorithm predicts
our test data more accurately than the Support Vector Machines (SVM) based kernels. And
also, the simulation results indicate that our proposed method has better Bit error rate
(BER) and SE.

Contents
摘要 i
Abstract iii
Contents v
INTRODUCTION
1 Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Research Motivation and Objective . . . . . . . . . . . . . . . . . . . . . . . 2
3 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.1 The Guideline for basic 5G Waveform . . . . . . . . . . . . . . . . . 4
3.2 Different waveform candidates for 5G and Beyond 5G . . . . . . . . . 5
3.3 Multi-carrier Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.3.1 Windowing . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.3.2 Subcarrier-Wise Filtering . . . . . . . . . . . . . . . . . . . 6
3.3.3 Subband-Wise Filtering . . . . . . . . . . . . . . . . . . . 7
3.4 Single-Carrier Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4 Proposed Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5 Contribution and Achievement . . . . . . . . . . . . . . . . . . . . . . . . . 9
6 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
BACKGROUNDS 10
1 Cyclic Prefix Orthogonal Frequency Division Multiplexing (CP-OFDM) . . . 10
2 Single-Carrier Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1 Cyclic Prefix DFT-s-OFDM . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Zero-tail-DFT-spread-OFDM (ZT-DFT-s-OFDM) . . . . . . . . . . . 13
2.3 Unique Word DFT-s-OFDM . . . . . . . . . . . . . . . . . . . . . . 14
3 Generalized Unique Word DFT-s-OFDM . . . . . . . . . . . . . . . . . . . . 15
Supervised Learning for Variable Guard Interval in OFDM 18
1 Data Generation for ML classification . . . . . . . . . . . . . . . . . . . . . . 19
2 Scaling of Delays by Data Formatting . . . . . . . . . . . . . . . . . . . . . . 22
3 Data Classification using Supervised Machine Learning Techniques . . . . . 23
3.1 Support Vector Machines (SVM) . . . . . . . . . . . . . . . . . . . . 25
3.2 k-Nearest Neighbors (KNN) . . . . . . . . . . . . . . . . . . . . . . . 26
4 Variable Guard Interval and Spectral Efficiency design . . . . . . . . . . . . 29
SIMULATION RESULTS 31
1 Simulation Environment and Parameters . . . . . . . . . . . . . . . . . . . . 32
2 Performance of Confusion Matrix and Accuracy of Classification Models: . . 32
3 Bit Error Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4 Spectral Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
CONCLUSIONS 41
Bibliography 45

List of Figures
1 Frame Structure of Cyclic Prefix, Internal guard interval and Variable guard
interval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1 Block diagram of CP-OFDM. (a) Transmitter. (b) Receiver [1] . . . . . . . . 11
2 Block diagram of CP-DFT-s-OFDM. (a) Transmitter. (b) Receiver [1] . . . . 12
3 An illustration of Internal guard interval and External guard interval [1] . . 13
4 Block diagram of Zero-Tail-DFT-s-OFDM. (a) Transmitter. (b) Receiver . . 14
5 Block diagram of Unique word-DFT-s-OFDM. (a) Transmitter. (b) Receiver [1] 15
6 Generalized UW DFT-S-OFDM transceiver system model [2] . . . . . . . . . 16
1 Flowchart for generating VGI . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2 (a) and (b) shows the data distribution before and after data formatting. . . 24
1 Classification performance of SVM linear kernel . . . . . . . . . . . . . . . . 33
2 Classification performance of SVM Polynomial kernel . . . . . . . . . . . . . 34
3 Classification performance of SVM Radial basis function kernel . . . . . . . 35
4 Classification performance of K-nearest neighbors . . . . . . . . . . . . . . . 36
5 Bite error rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6 Spectral efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

List of Tables
1 Comparison of Communication Networks [3] . . . . . . . . . . . . . . . . . . 1
1 Parameters of Tapped Delay Line Models [4] . . . . . . . . . . . . . . . . . . 20
1 The simulation parameters for waveform evaluation. . . . . . . . . . . . . . . 31
2 Different waveforms and its guard period . . . . . . . . . . . . . . . . . . . 38
3 The parameter for SE calculation . . . . . . . . . . . . . . . . . . . . . . . . 40
                                

[1] A. F. Demir, M. Elkourdi, M. Ibrahim, and H. Arslan, Waveform Design for 5G and
Beyond, pp. 51–76. 09 2018.
[2] S. Wei and J. 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, May 2020.
[3] K. S. Sankaran, G. Ramprabu, and V. Prakash, “Importance of fifth generation wireless
systems,” Mobile Computing, 2020.
[4] 3GPP, “Study on channel model for frequency spectrum above 6 GHz,” Technicl Specification (TS) 38.900, v14.2.0, Release 14, June. 2017.
[5] A. Chehri, P. Fortier, and P. M. Tardif, “Large-scale fading and time dispersion parameters of uwb channel in underground mines,” 2008.
[6] H. G. Myung, “Introduction to single carrier fdma,” in 2007 15th European Signal
Processing Conference, pp. 2144–2148, 2007.
[7] G. Berardinelli, K. I. Pedersen, T. B. Sorensen, and P. Mogensen, “Generalized DFTSpread-OFDM as 5G Waveform,” IEEE Communications Magazine, vol. 54, pp. 99–105,
November 2016.
[8] 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,
Dec 2015.
[9] Q. Incorporated, “Waveform Candidates,” 3GPP TSG-RAN WG1, Busan, Korea, April
11 - 15 2016.
[10] A. Sahin and H. Arslan, “Edge windowing for ofdm based systems,” IEEE Communications Letters, vol. 15, no. 11, pp. 1208–1211, 2011.
[11] B. Farhang-Boroujeny, “Ofdm versus filter bank multicarrier,” IEEE Signal Processing
Magazine, vol. 28, no. 3, pp. 92–112, 2011.
[12] N. Michailow, M. Matthé, I. S. Gaspar, A. N. Caldevilla, L. L. Mendes, A. Festag, and
G. Fettweis, “Generalized frequency division multiplexing for 5th generation cellular
networks,” IEEE Transactions on Communications, vol. 62, no. 9, pp. 3045–3061, 2014.
[13] V. Vakilian, T. Wild, F. Schaich, S. ten Brink, and J. Frigon, “Universal-filtered multicarrier technique for wireless systems beyond lte,” in 2013 IEEE Globecom Workshops
(GC Wkshps), pp. 223–228, 2013.
[14] X. Zhang, L. Chen, J. Qiu, and J. Abdoli, “On the waveform for 5g,” IEEE Communications Magazine, vol. 54, no. 11, pp. 74–80, 2016.
[15] A. Sahin, R. Yang, E. Bala, M. C. Beluri, and R. L. Olesen, “Flexible dft-s-ofdm:
Solutions and challenges,” IEEE Communications Magazine, vol. 54, no. 11, pp. 106–
112, 2016.
[16] G. Berardinelli, F. M. L. 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, Dec 2013.
[17] C. Gitau, “Classification in supervised machine learning: All you need to know!,” April 13 2018.
[18] A. Goel and S. K. Srivastava, “Role of kernel parameters in performance evaluation of svm,” in 2016 Second International Conference on Computational Intelligence Commu-nication Technology (CICT), pp. 166–169, Feb 2016.
[19] O. Harrison, “Machine learning basics with the k-nearest neighbors algorithm,” Sept 11 2018.
[20] N. Sharma, “Importance of distance metrics in machine learning modelling,” Jan 13 2019.
[21] 3GPP, “User Equipment (UE) radio transmission and reception; Part 1: Range 1 Stan-dalone,” Technicl Specification (TS) 38.101, V15.2.0, Sec. 6.5.2, Jul. 2018.
[22] 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, Dec 2015.

簡易檢索 / 詳目顯示

相關論文