一种基于稀疏自编码器的WiFi指纹库更新方法  

A WiFi fingerprint database update method based on sparse auto-encoder

作　　者：毛永毅 王志祥 MAO Yongyi;WANG Zhixiang(School of Electronic Engineering,Xi'an University of Posts and Telecommunications,Xi'an 710199,China)

机构地区：[1]西安邮电大学电子工程学院,西安710199

出　　处：《导航定位学报》2022年第3期137-144,共8页Journal of Navigation and Positioning

基　　金：陕西省工业攻关项目(2020GY-001)。

摘　　要：在基于无线保真(WiFi)指纹的室内定位系统中,指纹库是否能及时更新对于精确定位有着至关重要影响。针对无线接入点(AP)移动所带来的定位误差过大的问题,提出了一种利用稀疏自动编码器(SAE)进行指纹库更新的方法。通过对众包数据进行稀疏自编码,提取众包数据中所包含的隐藏特征用于原始指纹库的重构,期间引入均方误差MSE,有效地识别了AP的移动并对原始指纹库进行更新。实验结果表明,本方法有效降低了由于AP移动所带来的定位误差,在移动2个AP、4个AP、6个AP的情况下,相较于原始指纹库的平均定位精度分别提高了66.1%,75.3%,72.4%,且适用于移动不同AP数量下指纹库的更新,为室内WiFi指纹库的更新提供了一种切实可行的方案。In an indoor positioning system based on Wireless Fidelity(WiFi)fingerprints,whether the fingerprint database can be updated in time has a crucial impact on precise positioning.Aiming at the problem of excessive positioning error caused by the movement of wireless AP(Access Point),a method to update the fingerprint database by using SAE(sparse auto-encoder)is proposed in this paper.By sparse auto-encoding crowdsourced data,the hidden features contained in crowdsourced data are extracted for reconstruction of the original fingerprint database,during this period,mean square error M_(SE) was introduced to effectively identify the AP's movement and update the original fingerprint database.Experimental results show that this method effectively reduces the positioning error caused by AP movement.In the case of moving 2 APs,4 APs,and 6 APs,compared to the original fingerprint library,the average positioning accuracy is improved by 66.1%,75.3%,72.4%,and are suitable for moving different numbers of APs,providing a practical solution for the update of the indoor WiFi fingerprint database.

关 键 词：室内定位 众包数据 指纹库更新 无线接入点 稀疏自动编码器 

分 类 号：P228[天文地球—大地测量学与测量工程]