面向上肢运动功能康复训练-测评的
单目视频人体动作感知与识别技术研究
Study on Human Motion Perception And Recognition Technology Bad On Monocular Video for Training- Evaluation of Upper Limb Motor Function
Rehabilitation
受益无穷
决战食神演员表学科专业：生物医学工程
研究生：王璐
指导教师：张力新研究员农业项目有哪些
天津大学精密仪器与光电子工程学院
二零一二年十二月
独创性声明
独立英语
本人声明所呈交的学位论文是本人在导师指导下进行的研究工作和取得的研究成果，除了文中特别加以标注和致谢之处外，论文中不包含其他人已经发表或撰写过的研究成果，也不包含为获得天津大学或其他教育机构的学位或证书而使用过的材料。与我一同工作的对本研究所做的任何贡献均已在论文中作了明确的说明并表示了谢意。
学位论文作者签名：签字日期：年月日
学位论文版权使用授权书
本学位论文作者完全了解天津大学有关保留、使用学位论文的规定。特授权天津大学可以将学位论文的全部或部分内容编入有关数据库进行检索，并采用影印、缩印或扫描等复制手段保存、汇编以供查阅和借阅。同意学校向国家有关部门或机构送交论文的复印件和磁盘。
鹿柴怎么读（保密的学位论文在解密后适用本授权说明）
学位论文作者签名：导师签名：
签字日期：年月日签字日期：年月日
摘要
我国残疾总人口约为8296万，其中肢体残疾患者以29%的比例高居首位。尤其近年来，因各类事故以及中风等疾病所导致的上肢运动功能障碍人数呈显著上升之势。上肢运动功能障碍严重影响患者的日常生活，降低患者的生活质量。除了常规的手术治疗之外，康复治疗是此类患者恢复运动功能必不可少的有效手段。目前传统的康复治疗多是在理疗师的人力帮助下进行单调乏味的重复性作业训练，其成本高、自主性差、且无法对康复效果进行实时准确的评价。
为此，本文设计了一种面向上肢运动功能康复训练-测评的单目视频人体动作感知与识别技术，旨在通过运动手部动作跟踪及静态手势识别的方式实现康复过程中的自动人-机交互控制，同时在线获取上肢关节角度信息用于康复评价。研究中首先利用单目摄像头采集带有标示颜色手套的被试在三种典型上肢关节运动模式（肩关节冠状面运动、肩关节矢状面运动、肘关节矢状面运动）的视频图像以及9种不同手势命令的视频图像，摄像头的高度以及距离人体的距离根据人体解剖学统计和摄像头透射模型确定。后续的运动手部目标跟踪采用具有较好鲁棒性以及时间复杂性较低的Camshift算法，通过跟踪获取手部的实时位置坐标信息。静态手势识别分别通过预处理（手势分割、边界提取、降采样等）获得手势的边界轮廓，提取傅里叶描述子、边界方向直方图、边界不变矩三种特征，利用模板匹配对270例手势图像样本进行训练，540例进行测试，并对结果进行了对比分析；而后在特征层上进行了融合，有效提高了整体识别效果（平均识别率为91.85%，最高识别率达100%），弥补了单一特征信息的不足。关节角度提取尝试了两种方法：基于Hough变换的直线检测和基于跟踪结果的向量方向角，丝瓜的功效与作用
文中利用VICON动作捕捉系统与两种方法进行了比对分析，三种运动模式各取800例图像样本，两种方法的角度平均误差分别为16.34°和3.34°，相关性分别为0.833和0.885，结果显示了基于跟踪结果的向量方向角方法在关节角度提取方面的技术优势。本文研究成果有望进一步应用于融合机器视觉与虚拟现实技术的上肢康复临床实用系统的研制与开发。
关键词：上肢功能康复运动跟踪手势识别傅里叶描述子边界方向直方图不变矩模板匹配关节角度提取
ABSTRACT
The number of disabled persons in China is about 82.96 million, and physically disabled patients are with the highest proportion of 29% in all patients with disabilities. Especially in recent years, the number of upper limb motor function disorder becau of the accidents as well as stroke and other dias is significant increa in the trend. Upper limb motor dysfunction impact on the patient’s life riously and reduce the patient’s quality of life. In addition to the basic surgical treatment, rehabilitation is an effective and absolutely necessary means to restore patient’s motor function. Traditional rehabilitation is that patients do some repetitive task training in the help of physical therapist with high cost, poor autonomy, at the same time, and no real-time and accurate evaluation of the rehabilitation.
马丁路德金演讲稿
This thesis propod a new technology about human motion perception and recognition bad on monocular video for training-evaluation of upper limb motor function rehabilitation. Human-machine interaction control was achieved automatically through movement tracking and static gesture recognition during the rehabilitation process and the upper joint angles was acquired online for rehabilitation evaluation. In this study, monocular camera was ud to acquire the video images from the subject with the glove of special color in three kind of typical upper limbs motion mode (shoulder joint coronal movement, shoulder joint sagittal plane motion, elbow joint sagittal plane motion), and the video image of nine gestures command. The height of the camera and its distance from the human body was determined according to the human anatomy and camera transmission model. The Camshift algorithm was ud for subquent movement hand target tracking , which had better robustness and lower time complexity. Tthe coordinates of the real-time location of the hand portion were obtained by tracking. Static gesture recognition process included pretreatment (gesture gmentation, edge detection, and down-sampling), extracting three characteristics of boundary contour, including Fourier descriptors/boundary orientation histogram and boundary invariant moment, and then, pattern recognition with template matching for 270 samples of gesture images for training and 540 samples of gesture images for test. The results of recognition were compared and after the characteristics fusion at feature layer, effectively improve the whole recognition effect (the average recognition rate is
91.85% and the highest recognition rate is 100%) to make up for the lack of single feature. Two methods were tested to extract joint angles including the line detection bad on the Hough transform and the direction of the vector angle bad on tracking results, and then the results of the two methods were compared with the results from VICON motion capture system. With 800 samples of images for each motion mode, average angle error of two methods were 16.34° and 3.34°, respectively, and the correlation were 0.833 and 0.885, respectively. Results showed the advantages of the direction of the vector angle bad on tracking results for joint angles extraction. The achievements in this thesis may further applied in the rearch and development of practical clinical system for the rehabilitation of upper limbs combing machine vision and virtual reality technique.
KEY WORDS: Rehabilitation of upper limb function, Motion tracking, Gesture recognition, Fourier descriptors, Border direction histogram, Invariant moments, Template matching, Extraction of joint angle
泡菜煎饼
