人体下肢运动光纤带感知方法研究

Research on Optic Fiber Perception of Human Motion for Exoskeleton Robot Control

摘要:

外骨骼机器人实质上是一种可穿戴机器人,它将人的智能与外部机械能量结合在一起,可以给人提供额外的动力或能力,增强人体机能。在军事、康复医疗等领域有着巨大的应用价值和广阔的市场前景,也是国内外竞相研究的热点。为实现下肢外骨骼机器人与人体协调运动,首先需要实时的感知人体下肢的运动,并对运动进行模式识别,进而才有可能达到对下肢外骨骼机器人的实时控制的目的。本文对用于下肢外骨骼机器人控制的人体下肢运动光纤感知方法进行研究,在对人体下肢运动步态进行分析研究的基础上,提出了利用光纤测量下肢关节角度、利用分形理论对测得的下肢角度数据进行特征提取,并采用支持向量机对步态特征向量进行分类的识别方法。实验结果表明,该方法具有较高的识别性能,能够分辨出人体行走、跑、上斜坡和下斜坡、下蹲和起立等6种运动模式,且当核半径为0.4,惩罚因子为45时,识别率可达95%。

Abstract:

The exoskeleton robot is a wearable robot that will combine human intelligence with external mechanical energy mechanical power devices, can give extra power or ability to enhance human performance. It is used in the field of military, rehabilitation medical etc as a research hotspot. In order to make lower extremity exoskeleton robot follow human motion, and to lower limb exoskeleton robot real-time control, real-time measurement of human lower limb movement and the pattern recognition of the motion mode are need to be done. According to the lower limb gait analysis, the knee angle measurement points of human lower limb movement perception is selected and the fiber angle sensor is selected to measure the knee angle. Using fractal dimension to characterize the motion characteristics of the lower limbs. The SVM is used for pattern recognition by the value of the fractal feature vectors, then the classification parameters were optimized by the way of network search and cross-validation algorithms. The fiber optic sensing system for lower limb movement can identify the lower limb movement into severe motions: walking, running, climbing the slope, going down the slope, squatting and standing up. The optimization results shows that the best recognition results can reach 95% when nuclear radius is 0.4 and the penalty factor is 45.