张小红, 罗科干, 陶贤露, 胡鑫, 刘万科. 一种基于穿戴式MEMS传感器状态识别的多部位PDR算法[J]. 武汉大学学报 ( 信息科学版), 2021, 46(12): 1791-1801. DOI: 10.13203/j.whugis20210474
引用本文: 张小红, 罗科干, 陶贤露, 胡鑫, 刘万科. 一种基于穿戴式MEMS传感器状态识别的多部位PDR算法[J]. 武汉大学学报 ( 信息科学版), 2021, 46(12): 1791-1801. DOI: 10.13203/j.whugis20210474
ZHANG Xiaohong, LUO Kegan, TAO Xianlu, HU Xin, LIU Wanke. A Multi-mouted PDR Algorithm Based on Wearable MEMS Sensors State Recognition[J]. Geomatics and Information Science of Wuhan University, 2021, 46(12): 1791-1801. DOI: 10.13203/j.whugis20210474
Citation: ZHANG Xiaohong, LUO Kegan, TAO Xianlu, HU Xin, LIU Wanke. A Multi-mouted PDR Algorithm Based on Wearable MEMS Sensors State Recognition[J]. Geomatics and Information Science of Wuhan University, 2021, 46(12): 1791-1801. DOI: 10.13203/j.whugis20210474

一种基于穿戴式MEMS传感器状态识别的多部位PDR算法

A Multi-mouted PDR Algorithm Based on Wearable MEMS Sensors State Recognition

  • 摘要: 随着位置服务(location based service, LBS)应用需求的日益增加以及多部位微机电系统(micro electro mechanical system, MEMS)导航传感器的广泛普及,行人航位推算(pedestrian dead reckoning, PDR)越来越受关注,成为行人导航研究中主流的技术之一。但是,低成本的MEMS传感器测量噪声大,PDR解算误差积累严重;且PDR算法的普适性差,不同穿戴位置的MEMS导航传感器约束条件的可用性差异明显。提出了一种基于穿戴式MEMS传感器状态识别的多部位PDR算法。首先,采用支持向量机(support vector machine, SVM)进行全监督训练,实现了静止状态及运动状态下手部、腿部、腰部、足部4种穿戴位置的准确识别;然后,分析了不同穿戴位置下PDR算法的适用性,根据适用性分析结果提出了多部位PDR的综合解算策略。实测结果表明,该方法能够动态、准确地实现穿戴式MEMS传感器的状态识别,正确率达97%以上;应用PDR综合解算策略后,足部PDR能够实现高精度解算,累计误差为0.74%,而其他位置(手部、腿部、腰部)解算效果得到显著改善,累计误差从识别前的6.76%~21.19%减小为2.92%~5.62%。

     

    Abstract:
      Objectives  With the increasing demand for location-based service(LBS) applications and the wide popularity of multi-mounted micro electro mechanical system(MEMS) navigation sensors, pedestrian dead reckoning (PDR) algorithm has attracted more and more attention and has become one of the mainstream algorithms in pedestrian navigation research. However, the low-cost MEMS sensor has high measurement noise and serious accumulation of PDR solution error. Moreover, the universality of PDR algorithm is poor, and the availability of constraints of MEMS navigation sensors with different wearing statte is significantly different.
      Methods  A multi-mounted PDR algorithm based on wearable MEMS sensor state recognition is proposed. Firstly, support vector machine is used for fully supervised training to realize the accurate recognition of five wearing modes (hand, leg, waist, foot and stationary state). Then the applicability of PDR algorithm in different wearing states is analyzed, and a comprehensive solution strategy of multi position PDR is proposed based on the applicability analysis results.
      Results  The measured results show that the wear recognition accuracy of MEMS sensor is more than 97%. The foot PDR can achieve high-precision solution, and the cumulative error is 0.74%, while the solution effect of other positions (hand, leg and waist) has been significantly improved, and the cumulative error has been reduced from 6.76%—21.19% before recognition to 2.92%—5.62% after recognition.
      Conclusions  Therefore, the proposed algorithm can dynamically and accurately realize the state recognition of wearable MEMS sensors. After applying the PDR comprehensive solution strategy, the solution accuracy of PDR is significantly improved.

     

/

返回文章
返回