正交分頻多工(Orthogonal frequency division multiplexing, OFDM)無線電與光纖整合系統(Radio over fiber)被認為是非常具有潛力的次世代通訊架構，透過強度直接調變，將無線電波訊號載上光纖進行長距離傳輸。RoF技術將升降頻等工作集中在中央機房做處理，大幅度降低基地台複雜度以及維修成本，有利於大量基地台建設。然而在升降頻的過程都是透過而外的震盪器。而這樣子的過程難免會因為收發端的震盪器頻率不一致而導致所謂的載波偏移(Carrier frequency offset, CFO)，這個問題會導致基頻訊號產生相位轉移進而導致誤碼率上升。迴路上也會有取樣率偏差(Sampling clock offset, SCO)。與CFO相同的是，SCO也會產生相位轉移
本篇論文過單層神經網路可以在25公里光纖傳輸下，將帶有10ppm SCO與100ppm CFO進行補償。而透過深度學習則可以在100ppm SCO與200ppm CFO進行補償。在25公里光纖傳輸後外加3公尺毫米波傳輸下，單層神經網路可以補償10ppm SCO與50ppm CFO同時存在的情況，而三層神經網路則可以補償到100ppm。

Orthogonal frequency division multiplexing(OFDM) with Radio-over-fiber is a promising technology in next generation wireless communication, for its broadband and directly modulation and demodulation in transmitter and receiver. Through centralizing the difficult work to central office like up-conversion and down-conversion, reduce the complexity at base station(BS) then the BS can be widely built. However, the frequency of local oscillator for up-conversion and down-conversion oscillator almost can’t perfect match. The mismatch called Carrier Frequency Offset(CFO). This issue cause the base band signal got a phase which may influence the decision of the bit. The other problem is the Sampling Clock Offset(SCO) caused by the inaccurate sampling clock. Like CFO, SCO also cause the phase rotation but little different in the frequency domain.
In this paper, the signal with 10ppm SCO and 100ppm CFO under 25km fiber transmission can be compensated by single layer Neural network. 100ppm SCO and 200ppm CFO by Deep learning. In the other case, the signal with 100ppm SCO and 50ppm CFO under 25km fiber and 3m millimeter wave transmission can be compensated by Neural network, 100ppm SCO and 100ppm CFO by deep learning.

摘要 I
ABSTRACT II
致謝 III
目錄 IV
圖列表 VII
第一章 緒論 1
1.1 前言 1
1.2 研究動機 3
1.3論文架構 4
2.1正交分頻多工 Orthogonal frequency-division multiplexing, OFDM 5
2.2無線電波與光纖整合 7
2.2.1 RoF傳送端 7
2.2.2 RoF接收端 9
2.3.1載波偏移 Carrier frequency offset 12
2.3.2 取樣頻率誤差 Sampling Clock Offset, SCO 15
第三章 分類器 18
3.1貝氏定理 Bayes' theorem 18
3.2 似然性Likelihood function 19
3.3最大似然估計 Maximum Likelihood Estimation, MLE 21
3.4邏輯分類 Logistic regression[13] [14] 25
3.5貝氏分類器 Bayes classifier 28
3.6分類器到神經網路 Classifier to Neural Network 30
第四章 人工神經網路 31
4.1 神經網路原理 31
4.2 人工神經網路架構 32
4.3 神經元與神經層 Neuron and layer 33
4.4 權重與偏壓 Weights and bias 34
4.5 激活函數 Activation function 35
4.6 損失函數 Loss function 41
4.7 梯度下降法 Gradient descent 46
第五章 OFDM訊號生成與實驗架構 49
5.1 OFDM訊號生成 49
5.2 產生CFO與SCO 52
5.3 實驗架構 53
5.4 神經網路架構 55
第六章 實驗結果與分析 59
6.1 25km Fiber 59
6.2 25km Fiber 3m mm-wave wireless 63
第七章 結論 67
參考文獻 68
 

參考文獻
[1] Z. Pi and F. Khan, "An introduction to millimeter-wave mobile broadband systems," IEEE Communications Magazine, vol. 49, no. 6, pp. 101-107, 2011, doi: 10.1109/MCOM.2011.5783993.
[2] R. v. Nee and R. Prasad, OFDM for wireless multimedia communications. Artech House, Inc., 2000.
[3] T. Pollet, M. Van Bladel, and M. Moeneclaey, "BER sensitivity of OFDM systems to carrier frequency offset and Wiener phase noise," IEEE Transactions on communications, vol. 43, no. 2/3/4, pp. 191-193, 1995.
[4] T. Pollet, P. Spruyt, and M. Moeneclaey, "The BER performance of OFDM systems using non-synchronized sampling," in 1994 IEEE GLOBECOM. Communications: The Global Bridge, 28 Nov.-2 Dec. 1994 1994, pp. 253-257 vol.1, doi: 10.1109/GLOCOM.1994.513417.
[5] K.-S. Oh and K. Jung, "GPU implementation of neural networks," Pattern Recognition, vol. 37, no. 6, pp. 1311-1314, 2004.
[6] H.-R. Chen, "Enhanced Independence among NOMA users through Machine Learning Classification for 5G Downlink IFoF Network," 2018.
[7] "Wikipedia " https://zh.wikipedia.org/wiki/%E6%AD%A3%E4%BA%A4%E9%A0%BB%E5%88%86%E5%A4%8D%E7%94%A8 (accessed.
[8] 买在隔壁. "MZM及EAM的原理即特性公式推导." https://wenku.baidu.com/view/efb1022a453610661ed9f49d (accessed.
[9] Wikipedia. "Bayes' theorem." https://zh.wikipedia.org/wiki/%E8%B4%9D%E5%8F%B6%E6%96%AF%E5%AE%9A%E7%90%86 (accessed.
[10] L. I. Perlovsky and M. M. McManus, "Maximum likelihood neural networks for sensor fusion and adaptive classification," Neural Networks, vol. 4, no. 1, pp. 89-102, 1991.
[11] Wikipedia. "似然函數 Likelihood function." https://en.wikipedia.org/wiki/Likelihood_function (accessed.
[12] Wikipedia. "最大似然估計maximum likelihood estimation." https://zh.wikipedia.org/wiki/%E6%9C%80%E5%A4%A7%E4%BC%BC%E7%84%B6%E4%BC%B0%E8%AE%A1 (accessed.
[13] T. Huang. "機器/統計學習: 邏輯回歸(Logistic regression)." https://medium.com/@chih.sheng.huang821/%E6%A9%9F%E5%99%A8-%E7%B5%B1%E8%A8%88%E5%AD%B8%E7%BF%92-%E7%BE%85%E5%90%89%E6%96%AF%E5%9B%9E%E6%AD%B8-logistic-regression-aff7a830fb5d (accessed.
[14] Wikipedia. "邏輯迴歸 Logistic regression." https://zh.wikipedia.org/wiki/%E9%82%8F%E8%BC%AF%E8%BF%B4%E6%AD%B8 (accessed.
[15] T. Huang. "機器/深度學習: 基礎介紹-損失函數(loss function)." https://medium.com/@chih.sheng.huang821/%E6%A9%9F%E5%99%A8-%E6%B7%B1%E5%BA%A6%E5%AD%B8%E7%BF%92-%E5%9F%BA%E7%A4%8E%E4%BB%8B%E7%B4%B9-%E6%90%8D%E5%A4%B1%E5%87%BD%E6%95%B8-loss-function-2dcac5ebb6cb (accessed.
[16] T. J. Rouphael, RF and Digital Signal Processing for Software-Defined Radio. Elsevier, 2009.