基于机器视觉的抓握状态模型及其适用性

摘要

For digitising the hand movement of patients with hand function dysfunctions during rehabilitation, we presented that to recognise the hand state with a grasp state model, and analysed the applicability and robustness of the model by experiments.First, we presented the hand grasp state model, including the grasping objects, the types of hand grasp and the determination process of hand grasping state.Subsequently, we tracked using Leap Motion the hand movements of five testees when they grasping the ARAT objects, and analysed the applicability of the grasp state model.Finally, we analysed the effect of grasping objects size, hand differences and grasping angels on the stability of the model with the parameters of dispersion of the model in experiment.Experimental results showed that the average relative standard deviation of model parameters in experiment was 0.637, and the grasp state model has good applicability and robustness.Using computer vision-based grasp state model to reorganise the hand state basically meets the requirements of high track accuracy and rapid processing speed.%为了实现手部功能障碍患者在抓握康复训练中手部状态的数字化，提出采用抓握状态模型对手部状态进行识别，并通过实验分析模型的适用性与鲁棒性。首先，提出一种手部抓握状态模型，抓握对象，手部抓握类型以及手部抓握状态判定的流程。然后，采用Leap Motion对5名受试者抓握ARAT（ Action Research Arm Test）标准物时的动作进行跟踪，分析抓握状态模型的适用性。最后，通过实验中抓握状态模型参数的离散度分析抓握对象尺寸、手部差异和抓握角度对抓握状态模型稳定性的影响。实验结果表明：实验中抓握状态模型参数的平均相对标准偏差为0．637，且该抓握状态模型具有良好的适用性和鲁棒性。采用基于机器视觉的抓握状态模型对手部状态识别基本满足对手部运动跟踪精度高、处理速度快等要求。