指紋室內定位(fingerprint indoor localization)方法透過在不同位置的參考點(reference point)，接收事先佈置在特定位置之信標節點(beacon node)的信標(beacon)訊號，構成信標指紋(beacon fingerprint)以進行室內定位，具有不錯的定位精準度。為了改善不同定位訊號接收設備與不同定位區域間指紋定位的精準度，通常需要重新獲取新設備與新定位區域的信標訊號來進行定位資料校正。本論文提出基於領域自適應(domain adaptation, DA)及基於遷移學習(transfer learning)之指紋室內定位方法來進行定位資料校正。基於領域自適應方法將對應原設備與原定位區域的源域(source domain)與對應新設備與新定位區域的目標域(target domain)的定位資料分佈混和，藉此找到源域與目標域共同的特徵分佈空間，來改善目標域的定位精準度。基於遷移學習方法使用具有對應源域權重的自動編碼器進行遷移學習，以重構出適合目標域使用的自動編碼器。本研究收集室內指紋定位資料進行實驗，針對不同實驗場景評估所提方法之效能，並與相關方法比較。實驗結果顯示，在不同的訊號接收裝置以及不同的定位區域下，本論文所提方法皆有相對良好的定位精準度。

The fingerprint indoor localization method has good localization accuracy. It performs indoor localization by collecting beacon signals of beacon nodes at different reference points to form beacon fingerprints, where beacon nodes are pre-deployed at specific locations. To improve the localization accuracy for new beacon signal receiving devices and new localization zones, it is usually necessary to perform localization data (i.e., fingerprint) calibration by re-collecting beacon signals. This thesis proposes a method based on domain adaptation and a method based on transfer learning for performing fingerprint calibration. The method based on domain adaptation mixes the fingerprint distributions of the source domain and the target domain to find common distribution spaces for improving localization accuracy, where the former corresponds to the original beacon signal receiving devices and the original localization zone, and the latter corresponds to new beacon signal receiving devices and/or new localization zones. In the method based on transfer learning, the autoencoder with weights corresponding to the source domain is used to reconstruct, by the transfer learning scheme, the autoencoder with weights suitable for the target domain. Indoor fingerprint localization data are collected for conducting experiments to evaluate and compare the performance of the proposed methods and related methods. The experimental results show that the proposed methods have relatively good localization accuracy for different signal receiving devices and different localization zones.

中文摘要 I
目錄 IV
圖目錄 VI
一、 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 2
1.3 論文結構 2
二、 背景知識 3
2.1定位法 3
2.1.1 多邊定位法 (Trilateration) 3
2.1.2 三角定位法 (Triangulation) 4
2.1.3 指紋定位法 (Fingerprint) 5
2.2 RSSI 介紹 6
2.3 訊號強度影響 6
2.4 深度學習介紹 9
2.4.1 深度神經網路 (Deep Neural Network, DNN) 10
2.4.2 卷積神經網路(Convolutional Neural Network, CNN) 11
2.4.3 傳遞演算法 (Propagation) 13
2.4.4 自動編碼器 (AutoEncoder) 13
2.5 領域自適應模型 14
2.6 遷移學習 (Transfer Learning) 15
2.6.1 深度遷移學習 (Deep Transfer Learning) 16
基於實例的深度遷移學習 (Instances-based deep transfer learning) 16
基於映射的深度遷移學習 (Mapping-based deep transfer learning) 17
基於網路的深度遷移學習 (Network-based deep transfer learning) 17
基於對抗的深度遷移學習 (Adversarial-based deep transfer learning) 18
2.7 深度洞察法(DeepInsight) 19
2.8 相似性度量 20
2.8.1 歐氏距離 20
2.8.2 曼哈頓距離 20
2.8.3 閔可夫斯基距離 21
2.9 相關研究 21
三、研究方法 22
3.1 研究流程 22
3.2 收集源域及目標域資料 22
3.3 領域自適應模型訓練流程 22
3.3.1 指紋資料前處理 23
3.3.2 自動編碼器模型訓練 24
3.3.3 源域座標預測模型訓練 25
3.3.4 領域自適應模型訓練 25
V
3.4 基於自動編碼器的遷移學習模型訓練流程 27
3.4.1 指紋資料前處理 28
3.4.2 自動編碼器模型訓練 28
3.4.3 相似性測量 29
四、實驗結果 30
4.1 實驗設備介紹 30
4.2 實驗場景介紹 31
4.3 實驗結果與比較 33
FPFE-C方法介紹 33
方法一(領域自適應模型) 33
方法二 (基於自動編碼器的遷移學習) 38
相關研究比較 50
五、結論與未來展望 51
參考文獻 52

[1] 藍牙。取自https://en.wikipedia.org/wiki/Bluetooth
[2] Faragher, R., & Harle, R. (2015). Location fingerprinting with bluetooth low energy beacons. IEEE journal on Selected Areas in Communications, 33(11), 2418-2428.
[3] Ji, M., Kim, J., Jeon, J., & Cho, Y. (2015, July). Analysis of positioning accuracy corresponding to the number of BLE beacons in indoor positioning system. In 2015 17th International Conference on Advanced Communication Technology (ICACT) (pp. 92-95). IEEE.
[4] Jiang, J. R., Subakti, H., & Liang, H. S. (2021). Fingerprint Feature Extraction for Indoor Localization. Sensors, 21(16), 5434.
[5] Zuo, Z.; Liu, L.; Zhang, L.; Fang, Y. Indoor Positioning Based on Bluetooth Low-Energy Beacons Adopting Graph Optimization. Sensors 2018, 18, 3736.
[6] Martins, P.; Abbasi, M.; Sá, F.; Celiclio, J.; Morgado, F.; Caldeira, F. Intelligent beacon location and fingerprinting. Procedia Comput. Sci. 2019, 151, 9–16.
[7] Subedi, S.; Gang, H.-S.; Ko, N.Y.; Hwang, S.-S.; Pyun, J.-Y. Improving Indoor Fingerprinting Positioning With Affinity Propagation Clustering and Weighted Centroid Fingerprint. IEEE Access 2019, 7, 31738–31750.
[8] Li, M.; Zhao, L.; Tan, D.; Tong, X. BLE Fingerprint Indoor Localization Algorithm Based on Eight-Neighborhood Template Matching. Sensors 2019, 19, 4859.
[9] Dinh, T.-M.T.; Duong, N.-S.; Sandrasegaran, K. Smartphone-Based Indoor Positioning Using BLE iBeacon and Reliable Lightweight Fingerprint Map. IEEE Sens. J. 2020, 20, 10283–10294.
[10] 梁惠淞，基於校正之藍牙指紋室內定位，中央大學碩士論文(2010)
[11] He, S., & Chan, S. H. G. (2015). Wi-Fi fingerprint-based indoor positioning: Recent advances and comparisons. IEEE Communications Surveys & Tutorials, 18(1), 466-490.
53
[12] 多邊定位法、三角定位法。取自https://gisgeography.com/trilateration-triangulation-gps/
[13] Xia, S., Liu, Y., Yuan, G., Zhu, M., & Wang, Z. (2017). Indoor fingerprint positioning based on Wi-Fi: An overview. ISPRS International Journal of Geo-Information, 6(5), 135.
[14] Raju, M., Oliveira, T., & Agrawal, D. P. (2012, June). A practical distance estimator through distributed RSSI/LQI processing—An experimental study. In 2012 IEEE international conference on communications (ICC) (pp. 6575-6579). IEEE.
[15] M.Varsamou and T.Antonakopoulos. "A bluetooth smart analyzer in iBeacon networks", in IEEE Consumer Electronics??? Berlin (ICCE-Berlin), 2014
[16] Dong, Q., & Dargie, W. (2012, August). Evaluation of the reliability of RSSI for indoor localization. In 2012 International Conference on Wireless Communications in Underground and Confined Areas (pp. 1-6). IEEE.
[17] Huang, X., Barralet, M., & Sharma, D. (2009, March). Accuracy of location identification with antenna polarization on RSSI. In Proceedings of the International MultiConference of Engineers and Computer Scientists (Vol. 1).
[18] 深度神經網路。取自https://chenhh.gitbooks.io/multiperiod_portfolio_optimization/content/ml/ann/
[19] 損失函數。取自https://analyticsindiamag.com/loss-functions-in-deep-learning-an-overview/
[20] CNN架構示意圖。取自
https://medium.com/techiepedia/binary-image-classifier-cnn-using-tensorflow-a3f5d6746697
[21] 卷積運作示意圖。取自
https://medium.com/jameslearningnote/%E8%B3%87%E6%96%99%E5%88%86%E6%9E%90-%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E7%AC%AC5-1%E8%AC%9B-%E5%8D%B7%E7%A9%8D%E7%A5%9E%E7%B6%93%E7%B6%B2%E7%B5%A1%E4%BB%8B%E7%B4%B9-convolutional-neural-network-4f8249d65d4f
[22] Hecht-Nielsen, R. (1992). Theory of the backpropagation neural network. In Neural networks for perception (pp. 65-93). Academic Press.
[23] Wang, W., Huang, Y., Wang, Y., & Wang, L. (2014). Generalized autoencoder: A neural network framework for dimensionality reduction. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops (pp. 490-497).
54
[24] 自動編碼器架構圖。取自https://www.jeremyjordan.me/autoencoders/
[25] Tan, C., Sun, F., Kong, T., Zhang, W., Yang, C., & Liu, C. (2018, October). A survey on deep transfer learning. In International conference on artificial neural networks (pp. 270-279). Springer, Cham.
[26] 遷移學習流程圖。取自https://zh-tw.coderbridge.com/series/d4b5a1a1565e4e7a9cd14618ffe6146f/posts/54584ea6d4c240aeb3b8ae4af3a0531a
[27] Ganin, Y., & Lempitsky, V. (2015, June). Unsupervised domain adaptation by backpropagation. In International conference on machine learning (pp. 1180-1189). PMLR.
[28] Sankaranarayanan, S., Balaji, Y., Castillo, C. D., & Chellappa, R. (2018). Generate to adapt: Aligning domains using generative adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 8503-8512).
[29] Sharma, A., Vans, E., Shigemizu, D., Boroevich, K. A., & Tsunoda, T. (2019). DeepInsight: A methodology to transform a non-image data to an image for convolution neural network architecture. Scientific reports, 9(1), 1-7.
[30] Yong, L. H., & Zhao, M. (2020). Indoor positioning based on hybrid domain transfer learning. IEEE Access, 8, 130527-130539.
[31] Jiang, J. R., Subakti, H., Chen, C. C., & Sakai, K. (2018, August). PINUS: Indoor Weighted Centroid Localization with Crowdsourced Calibration. In International Conference on Parallel and Distributed Computing: Applications and Technologies (pp. 433-443). Springer, Singapore.
[32] Sun, Z., Chen, Y., Qi, J., & Liu, J. (2008, December). Adaptive localization through transfer learning in indoor Wi-Fi environment. In 2008 Seventh International Conference on Machine Learning and Applications (pp. 331-336). IEEE.
[33] Wang, G., Abbasi, A., & Liu, H. (2021, January). WiFi-based environment adaptive positioning with transferable fingerprint features. In 2021 IEEE 11th Annual Computing and Communication Workshop and Conference (CCWC) (pp. 0123-0128). IEEE.
[34] Mean-Square Error。取自https://towardsdatascience.com/understanding-the-3-most-common-loss-functions-for-machine-learning-regression-23e0ef3e14d3
[35] Seekcy iBeacon。取自http://www.seekcy.com/
[36] Van der Maaten, L., & Hinton, G. (2008). Visualizing data using t-SNE. Journal of machine learning research, 9(11).