English  |  正體中文  |  简体中文  |  Items with full text/Total items : 26988/38789
Visitors : 2317470      Online Users : 39
RC Version 4.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Adv. Search

Please use this identifier to cite or link to this item: http://ntour.ntou.edu.tw:8080/ir/handle/987654321/49301

Title: 語意式3D運動分析方法設計
On the Design of Semantic 3D Sports Analysis
Authors: Yu, Wan-Hsuan
Contributors: NTOU:Department of Computer Science and Engineering
Keywords: 動態捕捉資訊;動態時間扭曲演算法;迭代最近點演算法;K-means分群演算法;字串核心;Kinect;支援向量分類器
Motion Capture data;DTW;ICP;K-means clustering;String kernel;Kinect;SVM
Date: 2015
Issue Date: 2018-08-22T06:56:26Z
Abstract: 運動,是每個人在生活當中不可缺少的一部分,隨著體感控制系統崛起,如Kinect,透過闖關遊戲來達到運動的目的,但無法以專業立場主動矯正學習者的動作。因此,本研究提出一個語意式的3D動作分析方法的動作評分系統,讓學習上的動作更正確。 本研究利用動態捕捉資訊(Motion Capture Data)與Kinect取得人體的3D骨架資訊。演算法首先利用著名的迭代最近點(Iterative Closest Point,ICP ) 演算法來計算兩人體3D骨架姿勢的相似度,ICP用以計算兩組3D骨架模型的幾何轉換參數,藉以對齊輸入模型的控制點,達到姿勢校準的目的。並根據每一個動作序列內之姿勢的3D骨架座標點資訊,我們利用群組(K-means)演算法進行動作序列降維,將每一個動作內的姿勢進行分群,並找出屬於該動作的關鍵姿勢,此時輸入動作序列將由分群後的關鍵姿勢序列取代,由於動作序列中會有許多連續且重複的關鍵姿勢,在本研究中,我們去除連續且重複的部分,只留下一組關鍵姿勢序列代表該動作。最後,在訓練階段,本系統利用動態時間扭曲(Dynamic Time Warping,DTW)演算法計算兩關鍵姿勢序列的差異,並據以訓練每一個動作類別的樣板序列和使用字串核心(string kernel)之支援向量分類器(Support Vector Machine,SVM)。SVM動作分類器可用於辨識輸入3D骨架序列的關鍵姿勢之集合的動作類型,再透過動作評分系統可以給予使用者分數評比及專業教練的姿勢差異回饋。 實驗結果顯示本論文所提出的演算法可獲得良好的分類辨識準確度,並且透過建構互動式的學習系統,提供使用者在任何地方、任何時間中都可以動作,除此以外,本系統提供類似虛擬教練之分數及行為差異細部回饋,讓虛擬學習模式能更具實體感。本研究,可廣泛應用於不同的運動領域及相關的產業上。
Sports are of crucial importance in our everyday life. The Somatosensory technology, such as Kinect, becomes very important due to its potential lecture the professional skills of a trainee through playing sports computer games. To achieve the goal, this thesis presents a new approach to constructing a semantic 3D sports analysis system. Both Motion Capture (Mocap) data and Kinect Motion Capture (KMC) devices are used to build up the test datasets of human actions which are represented as 3D skeleton sequences. The cores of the system combines three modules. Firstly, Iterative Closest Point (ICP) is applied to the similarity calculation between two 3D skeletons by registering their joint points. Also, ICP computes the geometrical transformation between two 3D skeletons, which are represented as sets of 3D points. Secondly, the well-known K-means clustering algorithm is employed for dimensionality reduction to find representative postures (key-postures) in order to represent the input gesture as a compact key-posture sequence. Finally, Dynamic Time Warping (DTW) is conducted to calculate the difference between two key-posture sequences, which is then used to construct the string kernel of Support Vector Machine (SVM). The SVM, in the recognition phase, is used as the classifier to recognize the input gesture. Through the proposed sports scoring system (SSS), the system is capable of qualifying the user’s skill level in a class-specific sport by providing part-by-part gesture analyses in details. In the experimental results, the accuracy and stability of the proposed method is verified to demonstrate the effectiveness of the system using two public datasets, i.e., HDM05 of Mocap and THETIS of KMC. And, the system allows users to experience virtually interactive training without space and time limitations. To simulate the physical experiences on sports learning, this system offers virtual coaches to analyze sport gestures of the user in details. The propose approach can be widely appldied in a variety of sports area and related industries.
URI: http://ethesys.lib.ntou.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=G0010157028.id
Appears in Collections:[資訊工程學系] 博碩士論文

Files in This Item:

File Description SizeFormat

All items in NTOUR are protected by copyright, with all rights reserved.


著作權政策宣告: 本網站之內容為國立臺灣海洋大學所收錄之機構典藏,無償提供學術研究與公眾教育等公益性使用,請合理使用本網站之內容,以尊重著作權人之權益。
網站維護: 海大圖資處 圖書系統組
DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback