一种改进的行人导航算法研究  

An improved pedestrian navigation algorithm

作　　者：刘宇[1] 向高林[1] 王伊冰 陈燕苹[1] 吕玲[1] 黄河明[1] 

机构地区：[1]重庆邮电大学光电信息感测与传输技术重庆市重点实验室,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2016年第2期233-238,共6页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家自然科学基金(51175535);国际联合研究中心科技平台与基地建设(cstc2014gjhz0038);重庆市研究生科研创新项目(cys15163)~~

摘　　要：目前大多行人导航系统的研究都以手持手机、绑在腰间等单一姿态为前提,并未考虑行人使用手机过程中同时存在接电话、放在上衣口袋等多种姿态的情况。分析了行人使用手机的4种常见姿态,基于行人航迹推算(pedestrian dead reckoning,PDR)算法设计了一种改进的行人导航算法。利用神经网络(back propagation,BP)模型对手机姿态进行智能识别,设计了一种新的步态检测算法来对多姿态下的步态进行有效检测,并将行人行走过程分为单一姿态与姿态切换过程,对2个过程的航向进行了修正。软件测试结果表明,多姿态下实现系统定位精度在20‰以内,达标率在85%以上,应用于导航系统能够提供准确、可靠、持续的位置信息。Nowadays the premise of most pedestrian navigation system is used by a single pose,such as hand-held,tied to the waist and so on,but the diversity ways in which people use smart phones are not taken into consideration,such as taking a phone call,putting the phone in the pocket at the same time. This paper analyzes the four poses by which pedestrians use smart phones and an improved pedestrian navigation algorithm based on the PDR. Firstly,the algorithm used the BP neural network model to identify the smartphone attitude. Secondly,a new gait step detection algorithm was designed to detect the step in four poses. Finally,the pedestrian walking process can be divided into single posture and posture switching process. This paper corrected the heading of the two walking process. The software test results showed that achieving the control rate of system positioning precision within 20‰ was above 85% for the posture. The navigation system can provide accurate,reliable and continuous location information by the improved algorithm.

关 键 词：行人导航定位 神经网络(BP) 姿态识别 步态检测 航向修正 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]