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

Abstract

Contents

I. INTRODUCTION 12

II. DIAGNOSIS OF ADHD AND RELATED WORKS 15

2.1. Symptoms of children with ADHD 15

2.2. Diagnostic process for ADHD 17

2.2.1. Interview for ADHD diagnosis 20

2.2.2. Rating scale for the diagnosis of ADHD 21

2.2.3. Objective measurement method for diagnosing ADHD 22

2.3. Technical Methods for Diagnosing ADHD 23

2.4. Treatment of ADHD 26

III. MATERIALS AND METHODS 27

3.1. A robot-led game for the screening diagnosis of ADHD 27

3.2. Robot to assist in screening diagnosis-Silbot-3 34

3.3. Participants and data acquisition environment 35

3.4. Sensor system for acquiring data from children during games 38

3.4.1. Multiple subjects tracking system 42

3.5. Feature extraction for predefined measurement items 48

3.5.1. Indicator A1-body movement during wait 50

3.5.2. Indicator A2-Take a seat during wait 52

3.5.3. Indicator A3-Leave the waiting area during wait 52

3.5.4. Indicator B1-Enter the game board before the start instruction 53

3.5.5. Indicator B2-Not playing the game after the start instruction 54

3.5.6. Indicator B3-The amount of time from the instruction to start until the child begins the game 55

3.5.7. Indicator B4-the accuracy of the child's path movement 55

3.5.8. Indicator B5-move before the movement confirmation sound 57

3.5.9. Indicator C1 to C4-child's response to stimuli 58

3.5.10. Indicator C5-time taken to respond to a stimulus 59

3.5.11. Indicator D1 and D2-total execution time and distance traveled 60

3.6. A feature selection method to measure the rank of the features 61

3.7. Analysis of Feature Differences in Each Group using Bhattacharyya Distance and Cross Correlation of Data Probability Distributions 64

3.7.1. Bhattacharyya Distance 64

3.7.2. Cross Correlation 65

3.8. Feedback Reflection Architecture for Reflecting Collective Intelligence by Clinicians 66

3.9. Deep Neural Networks Model for Performance Validation 70

IV. EXPERIMENTAL RESULTS 73

4.1. Experiment 1: Improving acquired data accuracy using multiple Sensors 75

4.2. Experiment 2: Improving ADHD Screening Diagnostic Tool Performance with Video Labeling Feedback 77

4.3. Experiment 3: Analysis of differences in indicators (features) for each group (Normal group, ADHD at risk group, ADHD group) 80

4.3.1. Analysis of each group indicators using Bhattacharyya distance and Cross coefficient 80

4.3.2. Analysis of each group indicators using ratio comparison 84

4.4. Experiment 4: Consistency verification of the performance of the ADHD screening diagnostic tool using a robot 85

4.5. Experiment 5: Comparison with other screening diagnosis tools 88

4.6. Experiment 6: ADHD classification performance by gender 90

V. CONCLUSION AND DISCUSSION 91

APPENDICES 93

Appendix A. Research Progress 93

Appendix A.1. IMU-Location sensor and robot-based fall detection 93

Appendix A.2. Daily activities classification using contextual information 96

국문 요약 99

Reference 100

List of Tables

Table I. ADHD Diagnostic Criteria 17

Table II. The currently used ADHD evaluation scales in Korea 21

Table III. Diagnostic accuracy for ADHD of previous studies 25

Table IV. The list of abnormal behavior categories-type of 1:1 game 29

Table V. The stimuli and paths presented for each level 31

Table VI. Play method of domino game 33

Table VII. The list of abnormal behavior categories-type of 1:3 game 33

Table VIII. Data acquisition site for each session 35

Table IX. Abnormal behavior indicators for screening diagnosis of ADHD 49

Table X. Extracted rank of the feature importance 62

Table XI. Feedback result of video labeling review 67

Table XII. Modification of index extraction algorithm according to feedback reflection 68

Table XIII. The research questions 73

Table XIV. Datasets of children who participated in the experiment 74

Table XV. The confusion matrix of features extracted before labeling validation 77

Table XVI. The confusion matrix of features extracted after labeling validation 77

Table XVII. Datasets of children who participated in the experiment (Training/Test) 78

Table XVIII. The performance of the screening diagnostic tool before and after feedback 78

Table XIX. The performance of the screening diagnostic tool using different models 79

Table XX. Bhattacharyya distance and Cross Coefficient of indicators 80

Table XXI. Proportion of path-failing children for each data group by game level 84

Table XXII. Datasets of children who participated in the experiment (All sessions) 86

Table XXIII. The performance of the screening diagnostic tool using different environment 87

Table XXIV. The performance of the exist screening diagnostic tool 88

Table XXV. The performance of the screening diagnostic tool using robot 89

Table XXVI. The performance of the screening diagnostic tool for boys 90

Table XXVII. The performance of the screening diagnostic tool for girls 90

Table XXVIII. The confusion matrix using only acceleration data 94

Table XXIX. The confusion matrix using double-check method 95

Table XXX. The result of fall detection 95

Table XXXI. The confusion matrix using only GRU model 98

Table XXXII. The confusion matrix using ensemble model 98

List of Figures

Figure 1. Schematic Diagram of this study 14

Figure 2. Characteristics of ADHD disorder 16

Figure 3. The diagnostic process for ADHD 19

Figure 4. An actual test example of "A robot-led game for the screening diagnosis of ADHD"-type of 1 versus 1 28

Figure 5. Stimuli presented during the game, Left: Friends (Scarecrow, Tin Woodman, Lion), Right: Witch 30

Figure 6. An actual test example of "A robot-led game for the screening diagnosis of ADHD"-type of 1 versus 3 32

Figure 7. Silbot-3 Specifications 34

Figure 8. The environment of session 1-for type of 1 versus 1 game 36

Figure 9. The environment of session 2 and 3-for type of 1:1 or 1:3 games 37

Figure 10. (A) Optitrack IR camera by Natural Point, (B) People wearing special suits with IR markers performing motion capture 38

Figure 11. Skeleton figure(A) and base figure(B) estimated using Openpose algorithm 40

Figure 12. Various RGB-D sensors, (A) Intel's Realsense, (B) Orbbec's ASTRA, (C) Microsoft's Kinect V2, (D) Microsoft's Kinect Azure 41

Figure 13. DBSCAN algorithm 43

Figure 14. DBSCAN-based skeleton data merging 43

Figure 15. Using the finished system to recognize people identified by colored vests and their gestures 47

Figure 16. The indicators of abnormal behavior 48

Figure 17. The indicator A1-body movement during wait 50

Figure 18. The indicator A3-leaving the waiting area while the robot demonstrating 52

Figure 19. The indicator B1-Enter the game board before the start instruction 53

Figure 20. The indicator B2-Not playing the game after the start instruction 54

Figure 21. The indicator B4-the accuracy of the child's path movement 55

Figure 22. The indicator B4_1 (correct path) and B4_3 (incorrect path) 56

Figure 23. The indicator B4_2 (incorrect path before correct path) 56

Figure 24. The indicator B4_5 (correct path before incorrect path) 57

Figure 25. The indicator C1 to C4-child's response to stimuli 58

Figure 26. The indicator C5-time taken to respond to a stimulus 59

Figure 27. Indicator D1 and D2-total execution time and distance traveled 60

Figure 28. The schematic diagram of the wrapper method 61

Figure 29. Extracted rank of all features by RFE 62

Figure 30. The schematic diagram of the Feedback Reflection Architecture 66

Figure 31. process of labeling validation and feedback 67

Figure 32. Feedback of video labeling review 68

Figure 33. Effect by reflecting the collective intelligence of clinicians 69

Figure 34. A schematic diagram of MLP 70

Figure 35. A schematic diagram of dropout (Left) Standard Neural Net, (Right) After applying dropout 71

Figure 36. The MLP model used in the experiment 72

Figure 37. Behaviors to verify the skeleton data accuracy of the sensor 75

Figure 38. Sensor accuracy 76

Figure 39. Data Acquisition Environment in Session 1 85

Figure 40. Data Acquisition Environment in Session 2 and 3 86

Figure 41. The fall detection process using the IMU-L sensor and the robot 93

Figure 42. The schematic diagram of proposed daily activities classifier 96

Figure 43. Proposed ensemble model 97

초록보기

주의력 결핍 과잉 행동 장애(ADHD)는 부주의(주의 산만), 과잉 행동 충동과 같은 증상을 나타내는 혼합 행동 장애로 유전적, 신경학적, 심리사회적 연관성이 있다. ADHD는 일반적으로 어린 나이에 나타나 청소년기와 성인기까지 지속되는 장애이며, 따라서 이 질환은 조기 진단과 조기 치료가 필요한 질환이다. 지금까지 기존의 ADHD 진단 방법은 학부모, 교사 등의 관찰자에 의해 평가되어 객관적인 평가를 반영하는데 한계가 있다. 본 연구에서는 이러한 한계를 극복하기 위해 로봇이 주도하는 게임을 하는 어린이의 행동량과 주의력을 객관적으로 측정할 수 있는 비접촉 센싱 시스템을 개발하였다. 또한 멀티 레이어 퍼셉트론 (Multi-Layer Perceptron, MLP)과 센서를 통해 얻은 데이터를 이용하여 ADHD 군, ADHD 위험군, 정상군으로 분류할 수 있는 분류기를 개발하였으며, 이 시스템이 ADHD 선별진단에 유용하게 활용될 수 있는지 검증하였다. 본 연구는 서울 시내 6개 초등학교 사이트에서 828명의 초등학생을 대상으로 연구하였으며, 이 중 ADHD 군, ADHD 위험군, 정상군을 임상의의 진단과 비교하여 94.81%의 정확도를 달성하였다. 또한, 민감도가 ADHD 및 ADHD 위험군에서 각각 97.06% 및 100%를 달성했다는 점에서도 큰 의미가 있다.

추천서가 (다양한 추천 자료를 만나보세요)

권호기사보기

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