由於近年來深度學習技術的發展及普及，生活中許多的研究與發明，漸漸朝向人工智慧的方向發展，逐漸影響人們的日常體驗與生活，不論是在工商、金融、治安甚至是軍事及教育等等所有都能看到相關應用的出現，根據台灣衛福部統計處資料，在2019年跌倒事故傷害而過世的人竟然位居排行第二，故居家照護等相關應用也成了AI技術的一個重要議題，而跌倒偵測便是此次論文的研究重點。
有別於市售的穿戴式裝置如蘋果手錶和項鍊式緊急通報跌倒偵測器，利用設備的陀螺移、三軸加速度計或ECG心電圖等技術來判斷，為了避免人員發生意外時未配戴裝置很引發憾事，我們參考了攝影機影像辨識的技術，在特定場域裝設裝置判斷人員有無跌倒狀況，但礙於隱私權問題，會讓人有所顧慮，所以我們選擇了在場域架設雷達裝置來發展我們的跌倒偵測系統。
透過頻率調變連續波雷達(FMCW)，收集其回傳的原始資料(Raw data)，進行計算，產出範圍都卜勒圖(Range Doppler Image)及長時間間格的都卜勒直方圖(Long Time Interval Range Doppler Histogram)，觀察其資料特徵，對圖片及影像進行資料分析及編輯，並撰寫輔助工具，完成資料的收集及標籤(Label)，最後則是設計觸發(Trigger)模型，辨識圖片距離及速度變化量明顯的圖形，結合根據資料型態所自定義的的演算法完成觸發的判斷，再將有觸發的情形結合都卜勒長方圖丟至下一層基於雙向長短時記憶循環神經網路(Bi-directional Long Short-Term Memory)模型所設計的深度學習模型來做最後跌倒情形的判斷，並設計的簡易的告警機制，完成了高達90%以上準確率的跌倒偵測系統模型。

Owing to the development and popularization of deep learning technology in recent years, many daily researches and inventions have gradually developed toward to artificial intelligence, which is affecting people's lives including business, finance, transportation, public security or even military and education, etc. According to the Statistics Department of the Ministry of Health and Welfare, the number of people who died in fall accidents in 2019 ranked second, which leads to home care and human status detection becoming to an important development issue for AI technology. In this case, fall detection will be the focus of our research in this paper
Different from commercially wearable devices such as Apple Watch and Necklace-type emergency notification fall detectors using the device’s gyroscopic shift, three-axis accelerometer or ECG electrocardiogram to detect. In order to avoid accidents when people are not wearing the device from happening, we refer to the camera image recognition technology and install a device in a specific fiddle to determine/detect whether the person has fallen. Yet, due to privacy reasons, we decided to install radar device in the field to develop our fall detection system.
Through frequency modulated continuous wave radar(FMCW), collecting returned raw data, calculating, generating range doppler image and long time interval range doppler histogram, observing its data characteristics, analyzing and editing images, and writing auxiliary tools and completing data collection and label. Finally, we designed the Trigger model to recognize the changes of range doppler image, combined with the algorithm defined by the data type to complete the judgment of the trigger, and then dropped the triggered situation and long time interval range doppler Histogram into the next layer based on the Bi-directional Long Short-Term Memory model. It is designed to make the final judgment of the fall situation and also with a simple alarm mechanism; which overall, reach over 90% on accuracy, recall, precision and F1score of the fall detection system model

[1] Chang, Shyr-Juh, et al. "Taiwan quality indicator project and hospital productivity growth." Omega 39.1 (2011): 14-22.
[2] Hussain, Faisal, et al. "Activity-aware fall detection and recognition based on wearable sensors." IEEE Sensors Journal 19.12 (2019): 4528-4536. Tsai, Tsung-Han, Chin-Wei Hsu, and Wei-Chung Wan. "Live demonstration: Vision-based real-time fall detection system on embedded system." 2020 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 2020.
[3] Tsai, Tsung-Han, Chin-Wei Hsu, and Wei-Chung Wan. "Live demonstration: Vision-based real-time fall detection system on embedded system." 2020 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 2020.
[4] Jimi, Keiji, H. Seto, and Akihiro Kajiwara. "Bathroom Monitoring with Fast-Chirp Modulation Millimeter-wave UWB Radar." 2020 IEEE Radio and Wireless Symposium (RWS). IEEE, 2020. Fukushima, Shinya, et al. "Human motion detection and extraction by using FM-CW range-Doppler radar." 2014 International Symposium on Antennas and Propagation Conference Proceedings. IEEE, 2014.
[5] Daim, Terence Jerome, and Razak Mohd Ali Lee. "IR-UWB Radar Sensor Based Object Distance Determination as a Preliminary Input for Human Motion Detection." 2020 IEEE International Conference on Automatic Control and Intelligent Systems (I2CACIS). IEEE, 2020.
[6] Hayashida, Akira, Vasily Moshnyaga, and Koji Hashimoto. "The use of thermal ir array sensor for indoor fall detection." 2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC). IEEE, 2017.Hayashida, Akira, Vasily Moshnyaga, and Koji Hashimoto. "The use of thermal ir array sensor for indoor fall detection." 2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC). IEEE, 2017.Chang, Shyr-Juh, et al. "Taiwan quality indicator project and hospital productivity growth." Omega 39.1 (2011): 14-22.

[7] A unique approach to frequency-modulated continuous-wave radar design. Charvat,Gregory Louis.Michigan State University. ProQuest Dissertations Publishing, 2003. 1417824.
[8] Yang, Tao, Jun Cao, and Yongxin Guo. "Placement selection of millimeter wave FMCW radar for indoor fall detection." 2018 IEEE MTT-S International Wireless Symposium (IWS). IEEE, 2018.
[9] Radar technology principles: https://www.radartutorial.eu/index.en.html
[10] Raman, Sanjay, and Gabriel M. Rebeiz. "Integrated millimeter-wave polarimetric radar receivers." Proceedings of the 1996 IEEE National Radar Conference. IEEE, 1996.
[11] Li, Zhaolong, and Ke Wu. "24GHz FMCW radar front-end system on substrate." 2007 IEEE Radio and Wireless Symposium. IEEE, 2007.
[12] Rao, Sandeep. "Introduction to mmWave sensing: FMCW radars." Texas Instruments (TI) mmWave Training Series (2017).
[13] Iovescu、Cesar 和 Sandeep Rao。“毫米波傳感器的基本原理。” 德州儀器(2017)：1-8。
[14] Rajab, Khalid Z., et al. "Multi-target tracking and activity classification with millimeter-wave radar." Applied Physics Letters 119.3 (2021): 034101.
[15] Katsi, V., I. Felekos, and I. Kallikazaros. "Christian Andreas Doppler: A legendary man inspired by the dazzling light of the stars." Hippokratia 17.2 (2013): 113.
[16] Quinlan, J. Ross. "Induction of decision trees." Machine learning 1.1 (1986): 81-106.
[17] Pal, Mahesh. "Random forest classifier for remote sensing classification." International journal of remote sensing 26.1 (2005): 217-222.
[18] Wang, Sun-Chong. "Artificial neural network." Interdisciplinary computing in java programming. Springer, Boston, MA, 2003. 81-100.
[19] Liu, Weibo, et al. "A survey of deep neural network architectures and their applications." Neurocomputing 234 (2017): 11-26.
[20] LeCun, Yann, et al. "Gradient-based learning applied to document recognition." Proceedings of the IEEE 86.11 (1998): 2278-2324.
[21] He, Kaiming, et al. "Deep residual learning for image recognition." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
[22] Medsker, Larry R., and L. C. Jain. "Recurrent neural networks." Design and Applications 5 (2001): 64-67..
[23] Recurrent Neural Networks and LSTM explained. purnasai gudikandula.Mar 28,2019. https://purnasaigudikandula.medium.com/recurrent-neural-networks-and-lstm-explained-7f51c7f6bbb9
[24] Hochreiter, Sepp, and Jürgen Schmidhuber. "Long short-term memory." Neural computation 9.8 (1997): 1735-1780.
[25] Zhou, Peng, et al. "Attention-based bidirectional long short-term memory networks for relation classification." Proceedings of the 54th annual meeting of he association for computational linguistics (volume 2: Short papers). 2016.
[26] Chen, Suting, et al. "Strong Spatiotemporal Radar Echo Nowcasting Combining 3DCNN and Bi-Directional Convolutional LSTM." Atmosphere 11.6 (2020): 569.
[27] J. Lien, N. Gillian, M. E. Karagozler, P. Amihood, C. Schwesig, E. Olson, H. Raja,and I. Poupyrev, “Soli: Ubiquitous gesture sensing with millimeter wave radar,”ACM Trans. Graph., vol. 35, July 2016.
[28] Brett, Matthew, et al. "Region of interest analysis using an SPM toolbox." 8th international conference on functional mapping of the human brain. Vol. 16. No. 2. 2002.
[29] Szegedy, Christian, et al. "Inception-v4, inception-resnet and the impact of residual connections on learning." Thirty-first AAAI conference on artificial intelligence. 2017.
[30] Pedamonti, Dabal. "Comparison of non-linear activation functions for deep neural networks on MNIST classification task." arXiv preprint arXiv:1804.02763 (2018).
[31] Baldi, Pierre, and Peter J. Sadowski. "Understanding dropout." Advances in neural information processing systems 26 (2013): 2814-2822.
[32] Hawkins, Douglas M. "The problem of overfitting." Journal of chemical information and computer sciences 44.1 (2004): 1-12.
[33] Hay, A. M. "The derivation of global estimates from a confusion matrix." International Journal of Remote Sensing 9.8 (1988): 1395-1398.