基于深度学习的室内定位系统设计与实现

设计并实现了一个基于深度学习的室内定位系统,该系统分为信息采集模块、地磁定位模块、深度神经网络定位模块及联合定位模块4个模块。首先利用智能手机传感器采集室内地磁信号,以地理位置为标签保存为位置指纹文件。通过粒子滤波算法计算得出用户地理位置。然后使用深度神经网络模型和当前扫描的数据预测行人位置,最后对用户位置进行校正,消除地磁定位的累计误差。将基于深度学习的室内定位方法在Android平台实现并进行相应测试。实验结果表明,改进后的地磁室内定位的平均误差为2.2m,较只使用地磁定位技术进行定位时,定位误差平均降低了2.3m,具有更高的定位精度。

Design and implementation of indoor positioning system based on deep learning

An indoor positioning system is designed and inplemented based on deep learning,which is divided into four modules:information acquisition module,geomagnetic positioning module,depth neural network positioning module and joint positioning module.Firstly,the smart phone sensors are used to collect indoor geomagnetic signals,and the geographic locations are used as the labels to save as the location fingerprint file.The geographical location of users is calculated by particle filter algorithm.Then the depth neural network model and the current scanning data are used to predict the position.Finally,the user’s position is corrected to eliminate the accumulated error of geomagnetic positioning.The indoor location method based on deep learning is implemented and tested on Android platform.The experimental results show that the average error of the new indoor geomagnetic positioning is only 2.2 meters,which is 2.3 meters lower than that of only using geomagnetic positioning technology.