本論文提出一種建立於感測器融合(Sensor Fusion)架構的人體姿勢模態辨識之非監督學習演算法。該架構包含小型穿戴式的生醫電子偵測儀器以及相應的機械學習分析技術。架構中利用輕量化的偵測器感測人體姿勢訊號，並使用雲端服務傳輸至個人電子載具。同時研發出一種機械學習分析技術進行數據的運算與分群，以達到人體姿態的識別與資料庫的建立。該方法架構於量子力學的密度泛函理論(Density Functional Theory)：利用Hamiltonian密度以及Lagrangian密度分別估計資料點機率密度函數 I 之群數以及資料邊界。除可藉由此大幅簡化系統的計算複雜度以及提升可靠度與效率，更可同時計算資料點間相互的強度與連結性，以尋找資料群的邊界進行分群。本技術針對長期實驗可用以建立個人資料數據庫：除了準確偵測與分析個別人體姿態，亦藉由資料庫更新使得各種姿態頻率的判定更加精確。在這種長期的生理訊號的分析下可以找出各種人體可能發生的病灶，例如脊椎側彎等。短期實驗可用以意外事故肇因分析以及提前的警訊通知，如老人摔倒、幼兒翻身等。除可節省大量人力、物力資源與數據資料的佔用，同時具備更高的可信度與客觀性，乃至於爾後技術的商業化。

The thesis proposes an unsupervised leaning algorithm for human posture pattern recognition constructed in the framework of Sensor Fusion. The framework included a small wearable biomedical electronic sensing device and a corresponding analytical technique of machine learning. By means of the proposed sensing device, the sensed electronic signals would be transmitted to the personal electronic carriers through the cloud service. Meanwhile, the proposed algorithm would analyze the cluster behavior so that the personal database and posture patterns would then be constructed wherein. The algorithm was based on the density functional theory in the Quantum Mechanics. Using the Hamiltonian density functional and the Lagrangian density functional, the cluster number and the corresponding boundary of each cluster of a specific data probability density function I can be estimated. The proposed algorithm can not only dramatically reduce the computational complexity and reinforce the reliability and the efficiency, but also estimate the significance and the connectivity within data points to find the boundaries of clusters for further data clustering. For long-term observations, the proposed technique can be used to construct a customized database for sensing and analyzing the personal posture patterns, and also to classify each posture frequency by updating the database. Under this scenario, the possible lesions might be specified, such as the scoliosis and so forth. For short-term observations, it can be used for analysis of accident events and alert notification, such as elderly falls, children turn around, and so forth. Therefore, the proposed technique can save unnecessary costs caused from human interventions, material resources, and the occupation of storage. It also has higher credibility and objectivity, and even for the further commercialization.

[1] D. Laney, “3D Data Management: Controlling Data Volume”, Velocity and Variety. “ META group ,file:949. Retrieved 6 Feb. 2001.
[2] http://zh-tw.kionix.com/, Kionix, Inc. is a manufacturer of MEMS inertial sensors.
[3] http://tw.renesas.com/, Renesas Electronics Corporation.
[4] Bas, E., Erdogmus, D., Draft, R. W., & Lichtman, J. W., “Local tracing of curvilinear structures in volumetric color images: Application to the Brainbow analysis”, J. Vis. Commun. Image R. 23, 1260-1271 (2012).
[5] Bas, E. & Erdogmus, D., “Piecewise linear cylinder models for 3-dimensional axon segmentation in Brainbow imagery”, International Symposium on Biomedical Imaging (ISBI), 1297-1300 (2010).
[6] Vasilkoski, Z. & Stepanyants, A., “Detection of the optimal neuron traces in confocal microscopy images”, J. Neurosci. Meth. 178, 197-204 (2009).
[7] Wang, Y., Narayanaswamy, A., Tsai, C.-L., & Roysam, B., “A broadly applicable 3-D neuron tracing method based on open-curve snake”, Neuroinform. 9, 193-217 (2011).
[8] Türetken, E., Benmansour, F., Andres, B., Pfister, H., & Fua, P., “Reconstructing loopy curvilinear structures using integer programming”, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1822-1829 (2013).
[9] Gala, R., Chapeton, J., Jitesh, J., Bhavsar, C., & Stepanyants, A., “Active learning of neuron morphology for accurate automated tracing of neurites”, FNANA 8, 1-14 (2014).
[10] Chothani, P., Mehta, V., & Stepanyants, A., “Automated tracing of neurites from light microscopy stacks of images”, Neuroinform. 9, 263-278 (2011).
[11] Türetken, E., González, G., Blum, C., & Fua, P., “Automated reconstruction of dendritic and axonal trees by global optimization with geometric priors”, Neuroinform. 9, 279-302 (2011).
[12] Zhang, Y., Chen, K., Baron, M., Teylan, M. A., Kim, Y., Song, Z., Greengard, P., & Wong, S. T. C., “A neurocomputational method for fully automated 3D dendritic spine detection and segmentation of medium-sized spiny neurons”, NeuronImage 50, 1472-1484 (2010).
[13] Peng, H., Long F., & Myers, G., “Automatic 3D neuron tracing using all-path pruning”, Bioinformatics 27, i239-i247 (2011).
[14] Rodriguez, A., Ehlenberger, D. B., Hof, P. R., & Wearne, S. L., “Three-dimensional neuron tracing by voxel scooping”, J. Neurosci. Meth. 184, 169-175 (2009).
[15] Hsu, Y., & Lu, H. H.-S., “Brainbow image segmentation using Bayesian sequential partitioning”, International Journal of Computer Information Systems and Control Engineering 7, 891-896 (2013).
[16] Shao, H.-C., Cheng, W.-Y., Chen, Y.-C., & Hwang, W.-L., “Colored multi-neuron image processing for segmenting and tracing neural circuits”, International Conference on Image Processing (ICIP), IEEE, 2025-2028 (2012).
[17] Wu, T.-Y., Juan, H.-H., Lu, H. H.-S., & Chiang, A.-S., ”A crosstalk tolerated neural segmentation methodology for brainbow images”, Proceeding of the 4th International Symposium on Applied Sciences In Biomedical and Communication Technologies (ACM ISABEL), 2011.
[18] http://www.cl.cam.ac.uk/, Computer Laboratory, Faculty of Computer Science and Technology, University of Cambridge, UK.
[19] http://www.invensense.com/, MEMS gyroscope, and motion processing technologies for consumer electronics, USA.
[20] Tsai HJ, Kuo TBJ, Lee GS, and Yang CCH,” Efficacy of paced breathing for insomnia: Enhances vagal activity and improves sleep quality”, Psychophysiology ,52:388-396(2015)
[21] Kuo TBJ, Hong CH, Hsieh IT, Lee GS, and Yang CCH, “Effects of cold exposure on autonomic changes during the last REM sleep transition and morning blood pressure surge in humans.”, Sleep Medicine, 15:986-997,(2014)
[22] Bonizzi P, Karel JMH, Meste O, and Peeters RLM, “SINGULAR SPECTRUM DECOMPOSITION: A NEW METHOD FOR TIME SERIES DECOMPOSITION”, Advances in Adaptive Data Analysis , Volume 6 , No. 4, 1450011.(2014)
[23] Costa M, Goldberger AL, and Peng CK, “Multiscale Entropy Analysis of Complex Physiologic Time Series”, Phys. Rev. Lett. 89, 068102,( Published 19 July 2002)
[24] Chin-Feng Lai and Yueh-Min Huang, “Adaptive Body Posture Analysis for Elderly-Falling Detection with Multisensors “,IEEE Computer Society ,1541-1672,(2010)
[25] O. Yurur, C.-H. Liu and W. Moreno “Unsupervised posture detection by smartphone accelerometer “, ELECTRONICS LETTERS, Vol. 49 No. 8,(2013)