A Noncontact Capacitive Sensing System for Recognizing Locomotion Modes of Transtibial Amputees
A Noncontact Capacitive Sensing System for Recognizing Locomotion Modes of Transtibial Amputees
复制标题
DOI：
10.1109/tbme.2014.2334316
复制
发表时间：
2014-12-01
影响因子：
4.6
通讯作者：
Wang, Qining
中科院分区：
文献类型：
--
作者：
Zheng, Enhao;Wang, Long;Wang, Qining
本文提出了一种非接触式电容传感系统（C-Sens）的运动模式识别的经胫骨截肢。C-Sens检测残肢和假肢之间物理距离的变化。传感前端内置于假肢接受腔中，而不接触皮肤。这种新颖的信号源提高了基于肌电图（EMG）信号的运动模式识别系统和基于从皮肤接触获得的电容信号的系统的可用性。为了评估C-Sens的性能，我们在六名截肢程度不同的经胫骨截肢者中进行了实验，他们从事六种常见的运动活动。不同的运动模式的电容信号是一致的和刻板的。重要的是，我们能够获得足够的信息信号，即使是严重肌肉萎缩的截肢者（即，缺乏用于模式分类的来自小腿肌肉的高质量EMG的截肢者）。与相位相关的二次分类器和选定的特征集，该系统能够作出连续的判断运动模式的平均准确率分别为96.3%和94.8%的摆动阶段和站立阶段（实验1）。此外，该系统能够实现令人满意的识别性能后，受试者重新戴上插座（实验2）。我们还验证了C-Sens在截肢者活动期间负重5 kg时对承重变化的鲁棒性（实验3）。这些结果表明，非接触式电容感测是能够规避的EMG系统的实际问题，而不牺牲性能，因此，它是有前途的自动识别人体运动意图控制动力假肢。
This paper presents a noncontact capacitive sensing system (C-Sens) for locomotion mode recognition of transtibial amputees. C-Sens detects changes in physical distance between the residual limb and the prosthesis. The sensing front ends are built into the prosthetic socket without contacting the skin. This novel signal source improves the usability of locomotion mode recognition systems based on electromyography (EMG) signals and systems based on capacitance signals obtained from skin contact. To evaluate the performance of C-Sens, we carried out experiments among six transtibial amputees with varying levels of amputation when they engaged in six common locomotive activities. The capacitance signals were consistent and stereotypical for different locomotion modes. Importantly, we were able to obtain sufficiently informative signals even for amputees with severe muscle atrophy (i.e., amputees lacking of quality EMG from shank muscles for mode classification). With phase-dependent quadratic classifier and selected feature set, the proposed system was capable of making continuous judgments about locomotion modes with an average accuracy of 96.3% and 94.8% for swing phase and stance phase, respectively (Experiment 1). Furthermore, the system was able to achieve satisfactory recognition performance after the subjects redonned the socket (Experiment 2). We also validated that C-Sens was robust to load bearing changes when amputees carried 5-kg weights during activities (Experiment 3). These results suggest that noncontact capacitive sensing is capable of circumventing practical problems of EMG systems without sacrificing performance and it is, thus, promising for automatic recognition of human motion intent for controlling powered prostheses.