腱驱动假手手指运动及力分析

为优化多自由度假手的机械结构,降低其复杂性以及提高其可靠性,采用腱驱动方法实现假手单手 指的独立运动．腱驱动假手手指由驱动腱、钢丝轮、3个指节、3个关节及关节扭簧等构成,关节扭簧实现关节 的预紧和腱松弛后关节的回复运动．对弹簧刚度进行优化设计,实现腱驱动的手指各关节的顺序弯曲及手指 的大抓握．分析了手指抓取时各指节所受接触力与钢丝轮驱动力矩之间的关系．建立了手指的动力学模型, 得到手指的动力学方程．数值计算与仿真分析比较证明了理论计算的正确性,手指运动及抓握实验证明了该 种结构手指的可行性．腱驱动的假手手指具有零件少、结构简单、可靠性高等优点,为多自由度假手的研制提 供了有利条件     

基于EMG的假手实时力跟踪控制

针对商业假手灵活性及智能化水平较低的问题,进行了具有力矩/位置感知的五指假手及其控制方法的研究.以人体前臂肌电信号（EMG）为控制源,以力传感器检测指尖法向力,通过SVR支持向量机（Support Vector Regression,SVR）构建了指尖力的预测模型;采用网格搜索法确定模型的参数,缩短了训练时间,提高了模型的泛化能力.通过在线预测,结合带有二阶微分跟踪器的模糊PID控制方法实现了指尖施力实时跟踪;速度快,超调小,鲁棒性强,提高了假手的控制水平.     

一种新颖的手部运动跟踪系统

目前已有的手部运动跟踪系统大多在一定程度上限制了人体运动自由.基于此,提出一种无线、可穿戴、无障碍的腕关节、指关节运动跟踪系统.在人体每个手指甲上粘贴一轻小永磁体,用以产生标示腕关节、指关节运动的信号;若干磁传感器置于手腕处的电子腕带上,作为标示信号(磁信号)检测器.当腕关节、指关节运动时,永磁体在各传感器所在位置处的合成磁场发生变化,传感器对该磁场信号进行测量,所检测到的磁场信号送入手部姿势估计器,估计器基于系统数学模型计算手部姿势,从而实现对手部运动的跟踪.     

Postural stability when walking and exposed to lateral oscillatory motion: benefits from hand supports

While walking on a treadmill, 20 subjects experienced lateral oscillations: frequencies from 0.5 to 2 Hz and velocities from 0.05 to 0.16 m s^? 1 rms. Postural stability was indicated by ratings of ‘discomfort or difficulty in walking’, the movement of the centre of pressure beneath the feet and lateral forces applied to a hand support. Hand support improved postural stability with all frequencies and all velocities of oscillatory motion: the lateral velocity of the centre of pressure reduced by 30–50% when using support throughout motion, by 20–30% when instructed to use the support only when required and by 15% during normal walking without oscillation. Improvements in stability, and the forces applied to the hand support, were independent of support height when used continuously throughout motion. When support was used only when required, subjects preferred to hold it 118–134 cm above the surface supporting the feet.     

Effect of different handgrip angles on work distribution during hand cycling at submaximal power levels

The effect of different handle angles on work distribution during hand cycling was determined. Able-bodied subjects performed hand cycling at 20% of maximum power level (mean (SD) power level: 90.0 (25.8) W) at a cadence of 70 rpm using handle angles of ±30°, ±15° and 0°. The handle angle had a significant effect on work during the pull down (p < 0.001) and lift up (p = 0.005) sector, whereby the highest work was performed with handle angles of +30° and ?15° respectively. The cycle sector had a significant effect on work (p < 0.001) and significantly (p = 0.002) higher work was performed in the pull down sector (25% higher than mean work over one cycle) as compared to the lift up sector (30% lower than mean work over one cycle). Therefore, a fixed handle angle of +30° is suggested to be optimal for power generation. The results of this study help to optimise the handbike–user interface. A more pronated handle angle compared to the one conventionally used was found to improve the performance of hand cycling and thereby the mobility of disabled people.     

快速的三维人手运动跟踪方法研究

三维人手运动跟踪是人机交互领域的一个重要研究方向.提出了一种新的基于模型的三维人手运动跟踪方法,该方法将层次优化嵌入到基于粒子滤波器的跟踪框架中,通过在隐状态空间中对粒子采样来提高粒子滤波器采样效率.首先,提出了采用低维隐状态来描述人手的配置状态,并根据人手的生理运动约束建立人手动态模型;其次,为提高粒子在隐状态空间的采样效率,提出了采用层次遗传优化来快速地在局部寻找好的粒子,并以此作为重要度采样函数修正粒子滤波的采样算法.实验结果表明,该方法可以在人手自遮挡存在时的复杂背景下快速地对人手运动进行跟踪.     

基于多任务学习CNN辅助Transformer的手部mesh重建

针对当前手势姿态估计算法对于未充分利用2D信息辅助3D手部Mesh重建问题,首次在手部Mask、2D热力图的基础上提出引入RGB图像的HOG特征图,通过多任务学习CNN的框架对2D信息进行特征提取,并针对手的拓扑结构进行信息融合增强。为了解决Transformer encoder中的隐层embeddings维度一致性以及参数过大问题,设计了一种新的 MLP（multi-layer perception）模块嵌入Transformer encoder之间,达到embeddings的数量渐进增加以及其维度渐进减少的目的,从而完成手部mesh精细化的预测。实验结果表明该CNN与Transformer混合的新框架在 FreiHAND 和 RHD 数据集均取得了较好效果。     

Vision‐based control architecture for human–robot hand‐over applications

Human–robot collaboration will be an essential part of the production processes in the factories of tomorrow. In this paper, a human–robot hand‐over control strategy is presented. Human and robot can be both giver and receiver. A path‐planning algorithm drives the robotic manipulator towards the hand of the human and permits to adapt the pose of the tool center point of the robot to the pose of the hand of the human worker. The movements of the operator are acquired with a multi 3D‐sensors setup so to avoid any possible occlusion related to the presence of the robot or other dynamic obstacles. Estimation of the predicted position of the hand is performed to reduce the waiting time of the operator during the hand‐over task. The hardware setup is described, and the results of experimental tests, conducted to verify the effectiveness of the control strategy, are presented and discussed.     

基于Kinect的动态手势识别研究

针对人机交互领域中基于视觉的传统动态手势识别方法准确率不高、易受不同强度光照影响等问题,对动态手势识别方法进行了研究;首先利用Kinect传感器采集的深度图像对手势进行分割,并基于矩和链码进行手势质心与手势轨迹特征的计算,再利用动态时间规整算法进行手势轨迹特征识别,最后将识别结果传输给六足机器人进行人机交互实验,实现了动态手势对六足机器人的控制;实验结果证明:该方法识别准确率最高可达97%,且不易受光照影响,具有较强的鲁棒性,同时也满足了人机交互需求。