본문 바로가기 주메뉴 바로가기
22대 대한민국 국회 국회정보길라잡이 국회의원검색
회원가입 로그인
HOME

정보검색

소장정보 검색

국회도서관 홈으로 정보검색 소장정보 검색
결과 내 검색 동의어 포함
결과 내 검색 동의어 포함

-

목차보기

Title Page

Contents

Preface 8

Abstract 9

Chapter 1. Introduction 10

1.1. Introduction 10

1.2. Contributions 12

Chapter 2. Related Literature Review 13

2.1. Introduction 13

2.2. Appearance-Based Approaches 13

2.2.1. Colors and Shapes 14

2.2.2. Hand Features 14

2.2.3. The Viola-Jones Algorithm 15

2.2.4. Scale Invariant Feature Transform (SIFT) Features 16

2.3. 3D Hand Model-Based Approaches 16

2.4. Statistic & Syntactic Approaches 17

Chapter 3. System Architecture 19

3.1. Selection of Hand Gestures 19

3.2. System Architecture 20

Chapter 4. Hand Detection, Tracking, and Classification 22

4.1. Cursor Motion 22

4.2. Haar-Like Features 23

4.3. AdaBoost Learning Algorithm 28

4.4. Data Collection 32

4.5. Training and Testing 34

4.6. A Parallel Cascades Structure 35

4.7. Real Time Performance 36

4.8. Technique for Accurate Recognition 39

4.8.1. Preprocessing 40

4.8.2. Implementation 40

Chapter 5. Conclusion 42

References 43

List of Tables

Table 2.1. Statistical and syntactic approaches. 17

Table 4.1.The performance of the trained cascades of classifiers 35

List of Figures

Figure 2.1. Hand tracking using the color cue. 14

Figure 2.2. The "Visual Panel" system. 15

Figure 2.3. The block diagram of 3D hand model-based approaches. 17

Figure 3.1. Various hand postures selected for system implementation 19

Figure 4.1. Generation of image contour from RGB image 22

Figure 4.2. Effect of Median filtering 23

Figure 4.3. The extended set of Haar-like features 25

Figure 4.4. The weight compensation for different Haar-like features 26

Figure 4.5. The concept of "Integral Image" 27

Figure 4.6. The boosting process of the AdaBoost learning algorithm 30

Figure 4.7. Detect positive sub-windows using the cascade classifiers 32

Figure 4.8. Positive Samples 33

Figure 4.9. A fraction of the negative samples used in the training process 34

Figure 4.10. The parallel cascades structure for hand posture classification 36

Figure 4.11. Snaps of the real time results 39

Figure 4.12. Results for finding center and circle 41

초록보기

Hand gestures can be used for natural and intuitive human-computer interaction. To achieve this goal, computers should be able to visually recognize hand gestures from video input.

This thesis proposes a new architecture to interact with computers accurately and easily. This thesis optimizes pre-existing techniques of hand gesture recognition for mouse operation. Mouse operation has two parts, movement of the cursor and other mouse events like right mouse button and left mouse button. Color is used as a robust feature to generate a contour of hand and centroid of the contour or the center of palm is extracted and used for movement of cursor. Hue, Saturation and Value (HSV) color space is used in this regard. For other mouse operations the system was trained on Haar-like features and the gesture were classified through Adaboost classifier to recognize the gestures.

The Haar-like features can effectively catch the appearance properties of the hand gestures. The AdaBoost learning algorithm can significantly speed up the performance and construct an accurate cascade of classifiers by combining a sequence of weak classifiers. To recognize different hand postures, a parallel cascades structure is implemented. This structure achieves real-time performance and high classification accuracy.

Each gesture was connected with each of the mouse operation. An additional technique was added to improve the gesture recognition rate. It counts the number of fingers in a gesture to accurately classify a gesture.

추천서가 (다양한 추천 자료를 만나보세요)

권호기사보기

권호기사 목록 테이블로 기사명, 저자명, 페이지, 원문, 기사목차 순으로 되어있습니다.
기사명 저자명 페이지 원문 기사목차