隨著電腦計算速度的提升，機器學習已成為近年非常熱門的技術，此技術是利用電腦透過大數據找到資料之間的相關性，分析並建立模型，使用過去資料做訓練，並對未來資料進行預測。例如2017年由 Google DeepMind 公司開發的人工智慧圍棋程式 Alpha Go，擊敗了中國圍棋世界冠軍 - 柯潔，這項突破性的發展，展現了機器學習的價值。
由於手機常會因為手握住的關係，而導致通訊品質下降，所以我們考慮增加手機的接收天線，從中挑選幾路效能最好的來使用，以達到提升通訊品質的效果；我們將結合多輸入多輸出系統與機器學習中的深度學習，利用不同的特徵值經過神經網路的訓練，判斷出目前終端機的手握模式，並推薦能使接收效能改善的天線位置，最後我們搭配終端驗證平台觀察其位元錯誤率的變化，我們證實透過辨識及推薦系統可以降低錯誤率，提升吞吐量。

With the rapid growth of the computer calculate ability, the machine learning becomes feasible and thus becomes a very popular research topic in the recent year. By using the computer to find the correlation from the big data, we can analyze and construct a model which is able to predict the data from the original data. For instance, the Alpha Go, which is an A.I. program developed from the GOOGLE DeepMind has defeated the world champion of the chess in 2017. From this significant event, we can infer that the potential market of the machine learning will draw great attention.
The communication quality of mobile phones degenerate because of the hand-covered. To address this issue, we propose to increase the receiver antenna of the mobile phone, and then select a few of them. We combine the deep learning in the machine learning and the MIMO system to identify different hand-covered models. In particular, we use different features through the training of Neural Network to obtain the mode of hand-covered, and recommend the position index of the receiver antenna. Finally, we conduct simulations through a measurement platform to observe the change of bit error rates (BERs), and verify that the identification and recommendation system can improve the BER performance.

論文審定書 i
誌謝 ii
摘要 iii
Abstract iv
圖次 vii
表次 ix
第 一 章 緒論 1
1.1 前言 1
1.2 論文架構 3
第 二 章 系統架構 4
2.1 通訊系統架構 4
2.2 MIMO系統 5
2.3 LLR 8
第 三 章 深度學習判斷手握模式 9
3.1 深度學習 9
3.1.1 介紹 9
3.1.2 神經網路的基本架構 10
3.2 平台介紹 16
3.3 利用NEURAL NETWORK 辨識手握方式 17
3.3.1 資料的環境介紹 17
3.3.2 訓練特徵值的項目 19
3.3.3 決定訓練的特徵值 21
3.3.4 MIMO終端機量測平台實測 27
第 四 章 LORD Detector 29
4.1 LORD DETECTOR 29
4.1.1 QR分解 29
4.1.2 Slice 30
4.1.3 SWAP 31
4.2 運算複雜度 34
第 五 章 結果討論與分析 35
5.1 深度學習結果分析 35
5.2 檢測器模擬圖分析 39
第 六 章 結論 41
參考文獻 42

[1] C. Zhang and P. C. Woodland, “DNN speaker adaptation using parameterised sigmoid and relu hidden activation functions,” in Proc. IEEE Int. Conf. Acoust. Speech Signal Process., Shanghai, China, Mar. 2016, pp. 5300–5304.
[2] Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, pp. 436–444, May 2015.
[3] Y. Huo, X. Dong, and W. Xu, “5G cellular user equipment: from theory to practical hardware design,” arXiv preprint arXiv: 1704.02540, Apr. 2017.
[4] M. P. Milan. (2016, Oct. 07). Samsung galaxy S7, the first 4×4 MIMO smartphone. [Online]. Available: http://cellularinsights.com/samsung-galaxy-s7-the-first-4x4-mimo-smartphone/
[5] H. Sarieddeen, M. M. Mansour, L. M. A. Jalloul, and A. Chehab, “Low-complexity MIMO detector with 1024-QAM,” in Proc. IEEE Global Conf. Signal and Inf. Process, Orlando, FL, USA, Dec. 2015, pp. 883–887.
[6] A. Tomasoni, M.Siti, M. Ferrari, and S.Bellini, “Low complexity, quasi-optimal MIMO detectors for iterative receivers,” IEEE Trans. Wireless Commun., vol. 9, no. 10, pp. 3166–3177, Oct. 2010.
[7] Massimiliano Siti and Michael P. Fitz, “A novel soft-output layered orthogonal lattice detector for multiple antenna communications,” in Proc. IEEE Int. Conf. Commun., Istanbul, Turkey, Jun. 2006, pp. 1686–1691.
[8] K. Werner, H. Asplund, B. Halvarsson, A. K. Kathrein, N. Jalden, and D. V. P. Figueiredo, “LTE-A field measurements: 8x8 MIMO and carrier aggregation,” in Proc. IEEE 77th Veh. Technol. Conf., Dresden, Germany, Jun. 2013, pp. 1–5.
[9] K. Werner, H. Asplund, D. V. P. Figueiredo, N. Jaldén, and B. Halvarsson, “LTE-advanced 8x8 MIMO measurements in an indoor scenario,” in Proc. Int. Society Asphalt Pavements Conf. Antennas and Propagation., Nagoya, Japan Apr. 2012, pp. 750–753.
[10] S. Zhang, P. Harris, A. Doufexi, A. Nix, and M. Beach, “Massive MIMO real-time channel measurements and theoretic TDD downlink throughput predictions,” in Proc. IEEE 27th Annual Int. Symp. Personal, Indoor, and Mobile Radio Communications, Valencia, Spain, Sep. 2016.
[11] D. Boppana and A. Batada. (2005, Feb. 17). How to create beam-forming smart antennas using FPGAS. [Online]. Available: http://www.embedded.com/design/connectivity/4006442/How-to-create-beam-forming-smart-antennas-using-FPGAS.