本篇論文為使用被動式都卜勒雷達架構搭配注入鎖定正交接收機進行手語句子辨識的雷達系統，雷達訊號源為環境中5.8 GHz Wi-Fi訊號，天線接收手勢反射的回波訊號經正交解調器降至基頻，再由數位擷取卡取樣並傳輸到電腦進行後續的訊號處理。訊號處理的方式則是將手勢回波產生的IQ訊號，利用IQ訊號之間的相關性合併為一輸入資訊用於深度學習，並使用格拉姆角場將擁有時間序列資訊的數據轉換為圖像且此圖像能夠同時包含空間與時間資訊，以減少神經網路的輸入資訊而能夠大幅縮短深度學習訓練的時間。深度學習的部分則是使用卷積神經網路配合長短期記憶網路，能夠分別針對圖像與時間特徵進行訓練，以達到較佳的訓練效果。最後能以較少的數據集進行訓練並獲得91.2%的準確率，達成即時辨識由8個台灣手語單詞組合成的句子的手語辨識系統。

This study aims to develop a sign language recognition system using an injection-locking-based passive Doppler radar. The radar signal source comes from the ambient 5.8 GHz Wi-Fi signal. In this system, the Wi-Fi echo signals from the hands making sign languages are received by the antennas and then down-converted to baseband by the quadrature demodulators. After sampling with a digital acquisition card, the subsequent signal processing of the baseband signals is done by a computer. The signal processing method utilizes the correlation between the baseband I and Q signals as the input information of a deep learning process that incorporates Gramian angular field (GAF) to convert time series data into an image with spatial and temporal information. By doing so, the input information and learning time can be reduced to a great extent. The deep learning process uses a convolutional neural network (CNN) and a long short-term memory network (LSTM) to extract the image and temporal features, respectively, of the established GAF. Consequently, the system can achieve a higher accuracy of 91.2% with fewer training data sets in recognizing the Taiwanese sign language sentences that are composed from eight words in real-time.

論文審定書 i
誌謝 ii
摘要 iii
Abstract iv
目錄 v
圖次 vi
表次 viii
第一章 序論 1
1.1 研究背景與動機 1
1.2 Wi-Fi簡介 2
1.3 章節規劃以及研究目標 3
第二章 實驗架構 4
2.1 前言 4
2.2 系統架構 4
2.2.1 硬體架構介紹 4
2.2.2 天線位置調整 7
2.2.3 天線角度的調整 10
第三章 預處理及深度學習 14
3.1深度學習 14
3.1.1 深度學習介紹 14
3.2 格拉姆角場 (Gramian Angular Field, GAF)介紹與調整 16
3.2.1 調整GAF 18
3.2.2 降低輸入數 21
3.2.3 正規化方式 22
3.2.4 IQ訊號的格拉姆角場轉換 23
3.2.5 CW與Wi-Fi訊號比較 28
3.2.6 最佳訓練結果 30
3.3 即時句子辨識系統 31
3.3.1 辨識流程 31
第四章 結論 33
參考文獻 34

[1] Y. Zhang, D. Jiang, A. Demosthenous, “Design of a CMOS Analog Front-End for Wearable A-Mode Ultrasound Hand Gesture Recognition,” in Proc. 2022 17th Conference on Ph.D Research in Microelectronics and Electronics (PRIME), pp. 185-188, 2022
[2] B. B. Atitallah et al., “Simultaneous pressure sensors monitoring system for hand gestures recognition,” in Proc. IEEE SENSORS, pp. 1-4, 2020.
[3] Y. Wu, D. Jiang, X. Liu, R. Bayford and A. Demosthenous, “A Human--Machine Interface Using Electrical Impedance Tomography for Hand Prosthesis Control,” IEEE Trans. Biomed. Circuits Syst., vol. 12, no. 6, pp. 1322-1333, Dec. 2018.
[4] I. Ding and Z. Wu, “Two user adaptation-derived features for biometrical classifications of user identity in 3D-sensor-based body gesture recognition applications,” IEEE Sensors J., vol. 19, no. 19, pp. 8432-8440, Oct. 2019.
[5] S. Skaria, A. Al-Hourani, M. Lech and R. J. Evans, “Hand-gesture recognition using two-antenna Doppler radar with deep convolutional neural networks,” IEEE Sensors J., vol. 19, no. 8, pp. 3041-3048, Apr. 2019.
[6] S. Kim, Hong G. Han, J. Kim, S. Lee and T. Kim, “A Hand Gesture Recognition Sensor Using Reflected Impulses,” IEEE Sensors J., vol. 10, pp. 2975-2976, 2017.
[7] 王奕超, “Wi-Fi、藍牙、ZigBee－無線通訊三強，誰主沉浮？,” 2022 [Online]. Available:https://www.ibtmag.com.tw/new_article_result.asp?secu_id=HCP011&search_security_id=24983
[8] 黃品瑜，基於Wi-Fi都卜勒感測技術之及時手語辨識，國立中山大學電機工程學系碩士論文，2020。
[9] 林家齊，5.8 GHz Wi-Fi 都卜勒雷達之實現與生理監測應用，國立中山大學電機工程學系碩士論文，2021。
[10] 黃昱齊，2.4 GHz 數位波束成型都卜勒雷達之實現與生理監測應用，國立中山大學電機工程學系碩士論文，2019。
[11] 蔡素娟, 戴浩一, 陳怡君, “台灣手語線上辭典第三版中文版, ”國立中正大學語言學研究所, 2015年[Online]. Available: http://tsl.ccu.edu.tw/web/browser.htm
[12] B. Rohrer, “How do Convolutional Neural Networks work?,” 2016 [Online].
Available: https://e2eml.school/how_convolutional_neural_networks_work.html
[13] C. Olah, “Understanding LSTM Networks, ” 2015 [Online].
Available: https://colah.github.io/posts/2015-08-Understanding-LSTMs/
[14] L. de Vitry , “Encoding time series as images,” Medium., 2018 [Online].
Available:https://medium.com/analytics-vidhya/encoding-time-series-as-images-b043becbdbf3
[15] 社團法人台灣手語翻譯協會, “台灣親子手語數位學習網,” 2021 [Online]. Available:https://babysignlanguagetaiwan.taslifamily.org/