簡易檢索 / 詳目顯示

回結果列表
研究生： 葉冠宏
Yeh, Kuan-Hung
論文名稱： 體智能系統 – 包含單/多視角的具信效度折返跑計時方案
Motor–Cognitive Integrated Task (MCIT) - Including Reliable and Valid Shuttle Run Timing Solutions with Single/Multi-Perspective Views
指導教授： 范銘彥
Fan, Philex
共同指導: 蔡家齊
Tsai, Chia-Chi
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 電機工程學系
Department of Electrical Engineering
論文出版年： 2026
畢業學年度： 114
語文別： 英文
論文頁數： 115
中文關鍵詞： 人體姿態估計 、認知–動作評估 、即時運動測量 、運動分析
外文關鍵詞： Human Pose Estimation, Cognitive–Motor Assessment, Real-Time Sports Measurement, Sports Analysis
相關次數： 點閱：3  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 傳統執行功能評估多採用實驗室認知作業，然而這類方法對於動態、動作導向情境的推廣性仍有限，主要原因在於其大多忽略了身體動作在情境認知理論中扮演的角色。因此，本研究提出一套基於攝影機與純視覺之體智能整合任務（Motor-Cognitive Integrated Task, MCIT）折返跑測量系統，以檢驗動態動作情境下的認知功能表現。
    本框架整合場域評估、系統部署、基於幀數之事件計時，以及離線校正流程，僅需單一攝影機、筆記型電腦與螢幕即可運作。系統透過二維人體姿態估計與幾何定義之事件區域，即時推估動作–認知計時指標，例如反應時間、動作時間與總反應時間。系統允許加入額外攝影機，並做到同步錄製與即時判讀計時，此額外資訊可用於三維骨架結果的離線校正。
    驗證分析包括跨場域一致性、相機捕捉到筆記型電腦的端到端延遲、碰撞區域定義誤差分析，以及參考標準比較。跨場域結果顯示，區塊層級總反應時間具有良好一致性，一致性相關係數為0.867（95% 信賴區間：0.811–0.908）。但仍存在輕微系統性場域差異，平均偏差為0.068秒（95% 一致性界限：-0.098至0.234秒）。系統端到端延遲中位數為42.23毫秒。經三維骨架結果校正後之即時系統達到35毫秒之平均絕對誤差。上述結果支持本研究所提出之框架在校正條件下，可作為跨場域動作–認知評估之實用且可部署的測量方案。

    Executive function is typically assessed using well-validated laboratory paradigms, yet their generalizability to dynamic, action-based contexts is limited, as they largely neglect the role of bodily action emphasized in embodied and situated cognition frameworks. Accordingly, this study presents a camera-based, vision-only framework for real-time, cross-site shuttle run assessment within a Motor-Cognitive Integrated Task (MCIT) designed to examine cognitive function in action.
    The baseline configuration of the framework uses a single camera, a laptop, and a display to integrate site evaluation, deployment setup, frame-based event timing, and post hoc correction. Markerless 2D pose estimation and geometrically defined event regions were used to derive motor–cognitive timing indices in real time (e.g., reaction time, motor time, and response time). The framework can also be extended to a multi-view configuration by incorporating additional cameras for recording, without interrupting real-time timing, thereby providing complementary visual information for offline correction using 3D skeletal reconstruction.
    Validation analyses included cross-site consistency, end-to-end latency, collision region definition error analysis, and reference-standard comparison. Cross-site results showed good concordance for block-level response time (Lin’s concordance correlation coefficient (Lin’s CCC) = 0.867, 95% CI: 0.811–0.908), although a small systematic between-site shift remained (mean bias = 0.068 s; 95% limits of agreement: -0.098 to 0.234 s). Median end-to-end latency was 42.23 ms. The corrected real-time system using 3D-pose-adjusted results achieved a mean absolute error (MAE) of 35 ms. These findings support the feasibility of the proposed framework as a practical and deployable measurement solution for motor–cognitive assessment across sites under calibrated conditions.

    摘 要 iv Abstract vi Content ix List of Tables xii List of Figures xiii 1. Introduction 1 1.1 Motivation 1 1.2 Thesis Contribution 4 1.3 Thesis Organization and Structure 6 2. Background and Related Work 7 2.1 2D Human Pose Estimation 7 2.1.1 MoveNet 8 2.1.2 ViTPose 9 2.2 Multi-view 3D Human Pose Reconstruction 9 2.2.1 Monocular 3D HPE 10 2.2.2 Model-based Multi-view 3D HPE 10 2.2.3 Triangulation-based Multi-view HPE 11 2.3 Cognitive Task Paradigms and Stimulus Design 12 2.3.1 Simple Choice Reaction Time Task 13 2.3.2 Stroop Task 13 2.3.3 Simon Task 14 2.3.4 Flanker Task 15 2.3.5 Posner Cueing Task 16 2.4 Statistical Agreement and Repeated-measures Analysis 17 2.4.1 Lin’s CCC 18 2.4.2 Repeated-measures Bland–Altman Analysis 19 2.4.3 Linear Mixed-effects Model 21 2.4.4 Deming regression 22 2.5 Related Timing Systems for Sport and Movement Assessment 23 2.5.1 Wearable and Magnetometer-based Timing Systems 24 2.5.2 Camera-based Barrier Timing for Sprint Assessment 25 2.5.3 Vision-based Timing and Performance Analysis in Swimming 26 3. Proposed Methodology 28 3.1 System 28 3.1.1 Hardware 28 3.1.2 Software Architecture and Real-time Pipeline 29 3.1.3 Timestamping Policy 36 3.1.4 Offline-replay Pathway and Post Hoc Re-analysis 36 3.2 Task Design and Measurement Rationale 38 3.2.1 Task Structure and Stimulus–response Mapping 38 3.2.2 Rationale for Movement Boundaries and Collision Regions 39 3.2.3 Definition of RT, MT, and RMT 40 3.2.4 Trial-level Outcome Classification 41 3.3 Pose Estimation and Temporal Filtering 41 3.4 Geometric Layout and Event Evaluation Parameters 43 3.4.1 Camera Model and Undistortion 43 3.4.2 3D Definition and Projection of Boundaries and Collision Regions 47 3.4.3 Height-dependent Central Region Rule 50 3.5 Calibration and Reproducibility Protocol 51 3.5.1 Standardized Setup and Fixed-layout Parameters 51 3.5.2 Imaging Consistency Check for Intra-laboratory Repeatability 52 3.5.3 Cross-site Deployment Procedure 53 3.6 System Characterization and Error Correction 54 3.6.1 Reference-standard Annotation 54 3.6.2 Collision Region Definition Error 56 3.6.3 End-to-end Latency 56 3.6.4 Practical Cross-site Uncertainty and Sensitivity Considerations 57 3.7 Full Experimental SOP and Operator Checklist 57 4. Validation Experiments and Results 59 4.1 Reliability and Validity 59 4.2 Same-site Distributional Consistency across Two Camera Configurations 61 4.3 Cross-site Consistency of MCIT 63 4.4 Sensitivity Analysis under Controlled Setup 65 4.4.1 End-to-end Latency 65 4.4.2 Collision Region Definition Error 67 4.4.3 Cross-site Camera-pose Perturbation Analysis 72 4.4.4 Minimum Field-of-View Requirement for System Deployment 74 4.5 Reference-standard Comparison and Output Correction 76 4.5.1 Real-time System Correction Model using Eye-adjusted 76 4.5.2 2D-Offline-replay Correction Model using Eye-adjusted 77 4.5.3 Real-time System Correction Model using 3D-pose-adjusted 79 4.5.4 Real-time System Correction Model using 3D-eye-adjusted 81 4.6 Validation of Correction Equations 83 4.6.1 Real-time System Correction using Eye-adjusted Validation 83 4.6.2 2D-Offline-replay Correction using Eye-adjusted Validation 84 4.6.3 Real-time System Correction using 3D-pose-adjusted Validation 85 4.6.4 Real-time System Correction using 3D-eye-adjusted Validation 86 4.7 Comparison of Deployment, Hardware, and Timing Characteristics 87 5. Conclusion 92 6. Future Work 94 References 96

    [1] A. Diamond, “Executive functions,” Annual Review of Psychology, vol. 64, pp. 135–168, 2013.
    [2] S. F. Ahmed, S. Tang, N. E. Waters, and P. Davis-Kean, “Executive function and academic achievement: Longitudinal relations from early childhood to adolescence,” Journal of Educational Psychology, vol. 111, no. 3, pp. 446–458, 2019.
    [3] A. Kalén, E. Bisagno, L. Musculus, M. Raab, A. Pérez-Ferreirós, A. M. Williams, D. Araújo, M. Lindwall, and A. Ivarsson, “The role of domain-specific and domain-general cognitive functions and skills in sports performance: A meta-analysis,” Psychological Bulletin, vol. 147, no. 12, pp. 1290–1308, 2021.
    [4] L. W. Barsalou, “Grounded cognition,” Annual Review of Psychology, vol. 59, pp. 617–645, 2008.
    [5] K. R. Ridderinkhof, S. A. Wylie, W. P. M. van den Wildenberg, T. R. Bashore Jr., M. W. van der Molen, “The arrow of time: Advancing insights into action control from the arrow version of the Eriksen flanker task,” Attention, Perception, & Psychophysics, vol. 83, no. 2, pp. 700–721, 2021.
    [6] G. Vicente-Rodríguez, J. P. Rey-López, J. R. Ruíz, D. Jiménez-Pavón, P. Bergman, D. Ciarapica, J. M. Heredia, D. Molnar, A. Gutierrez, L. A. Moreno, and F. B. Ortega, “Interrater reliability and time measurement validity of speed-agility field tests in adolescents,” The Journal of Strength and Conditioning Research, vol. 25, no. 7, pp. 2059–2063, 2011.
    [7] C. Pérez-Chirinos Buxadé, B. Fernández-Valdés, M. Morral-Yepes, S. Tuyà Viñas, J. M. Padullés Riu, and G. Moras Feliu, “Validity of a magnet-based timing system using the magnetometer built into an IMU,” Sensors, vol. 21, no. 17, Art. no. 5773, 2021.
    [8] C. W. Bond, E. M. Willaert, and B. C. Noonan, “Comparison of three timing systems: Reliability and best practice recommendations in timing short-duration sprints,” The Journal of Strength and Conditioning Research, vol. 31, no. 4, pp. 1062–1071, 2017.
    [9] Noitom, “PERCEPTION NEURON 3,” FCC report. [Online]. Available: https://fcc.report/FCC-ID/2ABTR-NTM-MCP05TC01/5218161.pdf. [Accessed: May 19, 2026].
    [10] T. Uhlmann, S. Bräuer, F. Zaumseil, and G. Brunnett, “A novel inexpensive camera-based photoelectric barrier system for accurate flying sprint time measurement,” Sensors, vol. 23, no. 17, Art. no. 7339, 2023.
    [11] B. Buscà, M. Quintana, and J. M. Padullés, “High-speed cameras in sport and exercise: Practical applications in sports training and performance analysis,” Aloma: Revista de Psicologia, Ciències de l’Educació i de l’Esport, vol. 34, no. 2, pp. 13–23, 2016.
    [12] L.-X. Lee, K.-H. Yeh, C.-C. Tsai, D. Moreau, S.-C. Kao, and C.-H. Wang, “Toward ecological validity: A psychometric evaluation of a motor cognitive task integrating executive function and physical performance,” presented at the 16th World Congress of the International Society of Sport Psychology, 2025.
    [13] C.-H. Wang, C.-T. Yang, D. Moreau, and N. G. Muggleton, “Motor expertise modulates neural oscillations and temporal dynamics of cognitive control,” NeuroImage, vol. 158, pp. 260–270, 2017.
    [14] TensorFlow Blog, “Next-generation pose detection with MoveNet and TensorFlow.js,” 2021. [Online]. Available: https://blog.tensorflow.org/2021/05/next-generation-pose-detection-with-movenet-and-tensorflowjs.html. [Accessed: May 19, 2026].
    [15] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2: Inverted residuals and linear bottlenecks,” in Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, pp. 4510–4520.
    [16] E. Alam, A. Sufian, P. Dutta, and M. Leo, “Real-time human fall detection using a lightweight pose estimation technique,” arXiv preprint arXiv:2401.01587, 2024.
    [17] R. E. Kalman, “A new approach to linear filtering and prediction problems,” Journal of Basic Engineering, vol. 82, no. 1, pp. 35–45, 1960.
    [18] Y. Pei, S. Biswas, D. S. Fussell, and K. Pingali, “An elementary introduction to Kalman filtering,” Communications of the ACM, vol. 62, no. 11, pp. 122–133, 2019.
    [19] Q. Shuai, C. Geng, Q. Fang, S. Peng, W. Shen, X. Zhou, and H. Bao, “Novel view synthesis of human interactions from sparse multi-view videos,” in Proc. ACM SIGGRAPH Conference Proceedings, 2022, Art. no. 57, pp. 1–10.
    [20] L. I.-K. Lin, “A concordance correlation coefficient to evaluate reproducibility,” Biometrics, vol. 45, no. 1, pp. 255–268, 1989.
    [21] NCSS, LLC, “Lin’s concordance correlation coefficient,” in PASS Sample Size Software, ch. 812.
    [22] J. M. Bland and D. G. Altman, “Agreement between methods of measurement with multiple observations per individual,” Journal of Biopharmaceutical Statistics, vol. 17, no. 4, pp. 571–582, 2007.
    [23] Clinical and Laboratory Standards Institute, Measurement Procedure Comparison and Bias Estimation Using Patient Samples; Approved Guideline—Third Edition, CLSI document EP09-A3. Wayne, PA, USA: Clinical and Laboratory Standards Institute, 2013.
    [24] N. M. Laird and J. H. Ware, “Random-effects models for longitudinal data,” Biometrics, vol. 38, no. 4, pp. 963–974, 1982.
    [25] D. Bates, M. Mächler, B. Bolker, and S. Walker, “Fitting linear mixed-effects models using lme4,” Journal of Statistical Software, vol. 67, no. 1, pp. 1–48, 2015.
    [26] R. F. Martin, “General Deming regression for estimating systematic bias and its confidence interval in method-comparison studies,” Clinical Chemistry, vol. 46, no. 1, pp. 100–104, 2000.
    [27] S. Jin, L. Xu, J. Xu, C. Wang, W. Liu, C. Qian, W. Ouyang, and P. Luo, “Whole-body human pose estimation in the wild,” in Computer Vision—European Conference on Computer Vision, 2020, pp. 196–214.
    [28] CRONOX, “CRONOX 3.0—Wireless timing gates for speed sports.” [Online]. Available: https://cronox-sports.com/product/cronox-3-0/. [Accessed: May 19, 2026].
    [29] Sport-Thieme, “Witty Lichtschranken-Zeitmesssystem ‘Wireless’.” [Online]. Available: https://www.sport-thieme.de/Leichtathletik/Zubeh%C3%B6r/Leistungsdiagnostik/art=3217102. [Accessed: May 19, 2026].
    [30] KS Sports, “Wireless timing system with stopwatch, motion sensor and light barrier set.” [Online]. Available: Kabellose Zeitmess-Anlage, Set-Angebot, CHF 1'634.00. [Accessed: May 19, 2026].
    [31] Photo Finish, “Photo Finish.” [Online]. Available: https://photofinish-app.com/en/. [Accessed: May 19, 2026].
    [32] LYNX, “What is fully automatic timing (FAT) for sports?” [Online]. Available: https://finishlynx.com/about-us/what-is-fully-automatic-timing/. [Accessed: May 19, 2026].
    [33] Microsoft, “Microsoft Modern Webcam.” [Online]. Available: https://www.amazon.com/Microsoft-8L3-00001-Modern-Webcam/dp/B095DQJ3Z2. [Accessed: May 19, 2026].
    [34] Y. I. Abdel-Aziz, H. M. Karara, and M. Hauck, “Direct linear transformation from comparator coordinates into object space coordinates in close-range photogrammetry,” Photogrammetric Engineering & Remote Sensing, vol. 81, no. 2, pp. 103–107, 2015.
    [35] D. Mehta, S. Sridhar, O. Sotnychenko, H. Rhodin, M. Shafiei, H.-P. Seidel, W. Xu, D. Casas, and C. Theobalt, “VNect: Real-time 3D human pose estimation with a single RGB camera,” ACM Transactions on Graphics, vol. 36, no. 4, Art. no. 44, pp. 1–14, 2017.
    [36] P. Chen and C. Tsai, “3DSA: Multi-view 3D human pose estimation with 3D space attention mechanisms,” in Computer Vision—European Conference on Computer Vision, 2024, pp. 323–339.
    [37] F. Bogo, A. Kanazawa, C. Lassner, P. Gehler, J. Romero, and M. J. Black, “Keep it SMPL: Automatic estimation of 3D human pose and shape from a single image,” in Computer Vision—European Conference on Computer Vision, 2016, pp. 561–578.
    [38] A. Kanazawa, M. J. Black, D. W. Jacobs, and J. Malik, “End-to-end recovery of human shape and pose,” in Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, pp. 7122–7131.
    [39] A. F. R. Nogueira, H. P. Oliveira, and L. F. Teixeira, “Markerless multi-view 3D human pose estimation: A survey,” Image and Vision Computing, vol. 158, Art. no. 105437, 2025.
    [40] H. Tu, C. Wang, and W. Zeng, “VoxelPose: Towards multi-camera 3D human pose estimation in wild environment,” in Computer Vision—European Conference on Computer Vision, 2020, pp. 197–212.
    [41] A. Kadkhodamohammadi and N. Padoy, “A generalizable approach for multi-view 3D human pose regression,” Machine Vision and Applications, vol. 32, no. 1, pp. 1–14, 2021.
    [42] B. Jiang, L. Hu, and S. Xia, “Probabilistic triangulation for uncalibrated multi-view 3D human pose estimation,” in Proc. IEEE/CVF International Conference on Computer Vision, 2023, pp. 14804–14814.
    [43] K. Bartol, D. Bojanić, T. Petković, and T. Pribanić, “Generalizable human pose triangulation,” in Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 11028–11037.
    [44] X. Wan, Z. Chen, and X. Zhao, “View consistency aware holistic triangulation for 3D human pose estimation,” arXiv preprint arXiv:2302.11301, 2023.
    [45] J. Dong, W. Jiang, Q. Huang, H. Bao, and X. Zhou, “Fast and robust multi-person 3D pose estimation from multiple views,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 10, pp. 6981–6992, 2021.
    [46] Y.-R. Chern, Y.-H. Chen, F.-S. Lin, H.-C. Lin, G.-T. Chen, C.-P. Chu, G. Machtsiras, T.-H. Huang, J.-J. J. Lien, and C.-H. Huang, “A butterfly stroke swimming recording and performance analysis system based on computer vision and machine learning,” Measurement, vol. 251, Art. no. 117171, 2025.
    [47] J. M. Bland and D. G. Altman, “Statistical methods for assessing agreement between two methods of clinical measurement,” The Lancet, vol. 1, no. 8476, pp. 307–310, 1986.
    [48] K.-P. L. Vu and R. W. Proctor, “Mixing compatible and incompatible mappings: Elimination, reduction, and enhancement of spatial compatibility effects,” The Quarterly Journal of Experimental Psychology Section A: Human Experimental Psychology, vol. 57, no. 3, pp. 539–556, 2004.
    [49] J. Jiang, Q. Zhang, and S. van Gaal, “EEG neural oscillatory dynamics reveal semantic and response conflict at different levels of conflict awareness,” Scientific Reports, vol. 5, Art. no. 12008, 2015.
    [50] R. S. Redden, M. D. Hilchey, S. Aslam, J. Ivanoff, and R. M. Klein, “Using speed and accuracy and the Simon effect to explore the output form of inhibition of return,” Vision, vol. 7, no. 1, Art. no. 25, 2023.

    QR CODE