臺灣博碩士論文加值系統

醫療技術高度發展並持續精進的生活環境下，世界各地逐漸步入「高齡化社會」，長期照護已成為現今社會不容忽視的議題之一，其中跌倒偵測的技術也備受關注。然而過去研究大多使用加速度感測器，一般年長者配戴穿戴式裝置的意願極低，導致應用情境受限；本研究利用深度影像辨識技術，可以保護受測者的隱私。目的在於幫助長照領域更便利地追蹤年長者的即時行為，降低意外發生卻未能即時處理的遺憾；本研究利用長短期記憶網路（Long Short-Term Memory Network, LSTM）辨識人體動作，並分析異常行為（如：跌倒偵測），亦可讓使用者決定是否進行模型再訓練；提供更為完善的醫療照護功能。本研究分別在兩種情境下進行測試：在實驗室中建立不同連續動作，並將其所蒐集到的資訊應用於實際長照場域。實驗結果顯示：在實驗室的控制情境下可以準確辨識出受測者的動作，整體精確率91.33%、召回率88.86%；在實際長照場域下則會由於遮擋物及不確定因素之影響，將降低動作辨識績效，精確率73.75%、召回率67.52%。

Since the highly developed medical technology and improve the quality of life, the world is gradually entering the "Aging Society". Long-term care (LTC) has become one of the important issues today, and the technology of fall detection is highly eye-catching especially. However, most of the past studies used acceleration sensors which result in the less willingness. In this study, the depth image recognition is used to protect the subject's privacy issues. To help the long-term organization to track the elder’s immediate behavior easier. Furthermore, it also reduces regret of accidents events happen without timely processing. This study tries to use the Long Short-Term Memory Network (LSTM) to human action recognition and analyze abnormal behavior, e.g., fall detection, also allows users to decide whether to retrain the model or not. Provide better functions of the field of medical care. This study tested for two scenarios: Establish different continuous actions in the laboratory, and apply the collected information in LTC field. The experiment results show that under the control of the laboratory can accurately recognize the subject's actions, the overall precision is 91.33% and the recall rate is 88.86%. The other case which in the LTC fields, the performance of action recognition will be reduced due to the influence of obstructions and uncertain factors. The precision is 73.75% and the recall rate is 67.52%.

摘要 i
Abstract ii
目錄 iii
表目錄 v
圖目錄 vi
壹、 緒論 1
1.1研究背景與動機 1
1.2研究目的 3
1.3論文架構 3
貳、 文獻探討 4
2.1目標檢測 4
2.2人體姿態識別 5
2.2.1長短期記憶網路（Long Short-Term Memory Network, LSTM） 7
2.3跌倒偵測 9
參、 研究架構與方法 11
3.1研究架構 11
3.2人體部位追蹤定位 13
3.3辨識人體動作 14
3.3.1邊界問題 15
3.4跌倒偵測 16
3.5正常及異常行為分析 17
3.6配對及追蹤多人行為 20
肆、 實驗結果 22
4.1實驗設備 22
4.2執行效能 22
4.3績效計算方式 23
4.4動作辨識實驗結果 24
4.4.1實驗室資料集 25
4.4.2實際場域資料集 27
4.5跌倒偵測實驗結果 29
4.6使用者介面 31
伍、 結論與未來展望 33
5.1研究結論 33
5.2未來展望 34
參考文獻 36

世界衛生組織【實況報導】。民107年1月16日，取自：https://www.who.int/zh/news-room/fact-sheets/detail/falls
Bian, Z.-P., Hou, J., Chau, L.-P., & Magnenat-Thalmann, N. (2014). Fall detection based on body part tracking using a depth camera. IEEE journal of biomedical and health informatics, 19(2), 430-439.
Cao, Z., Simon, T., Wei, S.-E., & Sheikh, Y. (2017). Realtime multi-person 2d pose estimation using part affinity fields. Paper presented at the Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.
Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555.
Giles, C. L., Kuhn, G. M., & Williams, R. J. (1994). Dynamic recurrent neural networks: Theory and applications. IEEE Transactions on Neural Networks, 5(2), 153-156.
Girshick, R. (2015). Fast r-cnn. Paper presented at the Proceedings of the IEEE international conference on computer vision.
Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition.
Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
Hsieh, J.-W., & Hsu, Y.-T. (2008). Boosted string representation and its application to video surveillance. Pattern Recognition, 41(10), 3078-3091.
Hsieh, J.-W., Hsu, Y.-T., Liao, H.-Y. M., & Chen, C.-C. (2008). Video-based human movement analysis and its application to surveillance systems. IEEE Transactions on Multimedia, 10(3), 372-384.
Huang, Z., Liu, Y., Fang, Y., & Horn, B. K. (2018). Video-based Fall Detection for Seniors with Human Pose Estimation. Paper presented at the 2018 4th International Conference on Universal Village (UV).
Jansen, B., Temmermans, F., & Deklerck, R. (2007). 3D human pose recognition for home monitoring of elderly. Paper presented at the 2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society.
Kalman, R. E. (1960). A new approach to linear filtering and prediction problems.
Karantonis, D. M., Narayanan, M. R., Mathie, M., Lovell, N. H., & Celler, B. G. (2006). Implementation of a real-time human movement classifier using a triaxial accelerometer for ambulatory monitoring. IEEE transactions on information technology in biomedicine, 10(1), 156-167.
Lee, T., & Mihailidis, A. (2005). An intelligent emergency response system: preliminary development and testing of automated fall detection. Journal of telemedicine and telecare, 11(4), 194-198.
Li, Q., Stankovic, J. A., Hanson, M. A., Barth, A. T., Lach, J., & Zhou, G. (2009). Accurate, Fast Fall Detection Using Gyroscopes and Accelerometer-Derived Posture Information. Paper presented at the BSN.
Liu, J., Shahroudy, A., Xu, D., & Wang, G. (2016). Spatio-temporal lstm with trust gates for 3d human action recognition. Paper presented at the European Conference on Computer Vision.
Liu, J., Wang, G., Hu, P., Duan, L.-Y., & Kot, A. C. (2017). Global context-aware attention LSTM networks for 3D action recognition. Paper presented at the Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.
Livingston, M. A., Sebastian, J., Ai, Z., & Decker, J. W. (2012). Performance measurements for the Microsoft Kinect skeleton. Paper presented at the 2012 IEEE Virtual Reality Workshops (VRW).
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition.
Redmon, J., & Farhadi, A. (2018). Yolov3: An incremental improvement. arXiv preprint arXiv:1804.02767.
Ren, S., He, K., Girshick, R., & Sun, J. (2015). Faster r-cnn: Towards real-time object detection with region proposal networks. Paper presented at the Advances in neural information processing systems.
Schuster, M., & Paliwal, K. K. (1997). Bidirectional recurrent neural networks. IEEE transactions on Signal Processing, 45(11), 2673-2681.
Stone, E. E., & Skubic, M. (2015). Fall detection in homes of older adults using the Microsoft Kinect. IEEE J Biomed Health Inform, 19(1), 290-301. doi:10.1109/JBHI.2014.2312180
Ullah, A., Ahmad, J., Muhammad, K., Sajjad, M., & Baik, S. W. (2017). Action recognition in video sequences using deep bi-directional LSTM with CNN features. IEEE Access, 6, 1155-1166.