人體資勢偵測模型研究

The Research of Human Body Posture Detection Model

蘇有(Yu Su)

《中華印刷科技年報》 2016 (2016/04) Pp. 337-358

摘要

本文透過processing程式語言結合Kinect來設計一個體感介面，建立了一套人體姿勢偵測模型，讓使用者在運動的時候，得以取得參數了解自己的姿勢狀態，對於人體運動的相關推展有所助益。在醫療方面，此模型的建立有助於患者能透過正確的復健保健姿勢來加速身體的復原。在體感遊戲設計方面，本文的設計可以測試出何種體感動作能更適合遊戲的運行，增加遊戲的可玩性，也能讓遊戲兼具運動的效果。在運動學習方面，如舞蹈或武術的練習上，能藉由人體姿勢偵測模型來發揮一定的學習成效。processing程式語言配合SimpleOpenNI程式庫來開發人體姿勢偵測模型淺顯易懂，而使用本文建議的動態時間校正(DTW)演算法來做姿勢比對，精確且有效率。從實驗結果分析中得知，採用多張參考樣本來與玩家動作執行比對，將增強玩家的動作穩定性，若僅使用單張參考樣本，玩家的動作正確性振幅將過大，較不利於人體姿勢偵測。此外，參考樣本動作的設計將影響人體姿勢比對的結果，因此，設計一些合適的人體動作會增加體感應用程式的精確度。

關鍵字

人體姿勢偵測 ； 動態時間校正

並列摘要

In this study, Kinect combined processing programming language, designed to establish a somatosensory interface detection model of human posture. User when in motion, you can get to understand their posture state parameter, related to the implementation of human motion to be helpful. In medical terms, through this model helps patients through proper rehabilitation posture, speed up the body's recovery. In the somatosensory game design, through this model, we can test out the game for that kind of action to run, increase the game's playability, but also make the game both sports results. In sports learning, such as dance or martial arts exercises, through this model, we can play a good study results. SimpleOpenNI library binding processing programming language, development of human posture detection model is very simple, and the use of dynamic time warping (DTW) algorithm to do the posture alignment, accurate and efficient. Learned from the experimental results, the use of more than one reference sample and compare the user's actions, user actions will increase the stability of detection. If you use only one reference sample, the user's action amplitude comparison result will be much of a case, it is not conducive to human posture detection. In addition, reference samples action will be designed to affect the outcome of the comparison of human posture, therefore, the design of some suitable reference samples will increase accuracy somatosensory app.

並列關鍵字

Human Body Posture Detection ； Dynamic Time Warping

參考文獻

DiFilippo, N.,Jouaneh, M.(2015).Characterization of Different Microsoft Kinect Sensor Models.IEEE SENSORS JOURNAL.15(8),4554-4564.

Lozano-Quilis, JA.,Gil-Gomez, H.,Gil-Gomez, JA.,Albiol-Perez, S.,Palacios, G.,Fardoum, HM.,Mashat, AS.(2013).Virtual Reality System for Multiple Sclerosis Rehabilitation using KINECT.Pervasive Computing Technologies for Healthcare (Pervasive Health), 2013 7th International Conference on.(Pervasive Computing Technologies for Healthcare (Pervasive Health), 2013 7th International Conference on).

Arici, T.,Celebi, S.,Aydin, AS.,Temiz, TT.(2013).Robust Gesture Recognition using Feature Pre-processing and Weighted Dynamic Time Warping.Multimedia Tools and Applications.

Li, N.,Dai, Y.,Wang, R.,Shao, Y.(2015).Study on Action Recognition Based on Kinect and Its Application in Rehabilitation Training.Big Data and Cloud Computing (BDCloud), 2015 IEEE Fifth International.(Big Data and Cloud Computing (BDCloud), 2015 IEEE Fifth International).

Ruan, X.,Tian, C.(2015).Dynamic Gesture Recognition based on Improved DTW Algorithm.Mechatronics and Automation (ICMA), 2015 IEEE International Conference.(Mechatronics and Automation (ICMA), 2015 IEEE International Conference).