GPS-Based Indoor/Outdoor Detection Scheme Using Machine Learning Techniques
GPS-Based Indoor/Outdoor Detection Scheme Using Machine Learning Techniques
复制标题
使用机器学习技术的基于 GPS 的室内/室外检测方案
DOI：
--
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Y. Jang
中科院分区：
文献类型：
--
作者：
Van Bui;N. Le;ThanhLuan Vu;V. Nguyen;Y. Jang
移动的通信中的最新进展要求室内/室外环境信息可用于各个应用和无线信号传输，以便改善干扰控制并服务于上层应用。在本文中，我们提出了一个方案，以确定室内/室外环境使用GPS信号结合机器学习分类技术。与基于接收信号强度指示（RSSI）的传统方案相比，该方案具有鲁棒性强、精度高、在室内和室外环境之间移动时运行平稳、易于实现和训练等优点。该方案利用移动的设备上的GPS传感器，将来自一定数量的GPS卫星的信息结合起来。然后，使用针对最佳性能进行优化的机器学习模型收集、预处理数据，并将其分类为室内或室外环境。GPS输入数据收集在国民大学地区，包括850个训练样本和170个测试样本。整体准确率达到97%。
Recent advances in mobile communication require that indoor/outdoor environment information be available for both individual applications and wireless signal transmission in order to improve interference control and serve upper-layer applications. In this paper, we present a scheme to identify the indoor/outdoor environment using GPS signals combined with machine learning classification techniques. Compared to traditional schemes, which are based on received signal strength indicator (RSSI), the proposed scheme promises a robust approach with high accuracy, smooth operation when moving between indoor and outdoor environments, as well as easy implementation and training. The proposed scheme combined information from a certain number of GPS satellites, using the GPS sensor on mobile devices. Then, data are collected, preprocessed, and classified as indoor or outdoor environment using a machine learning model that is optimized for the best performance. The GPS input data were collected in the Kookmin University area and included 850 training samples and 170 test samples. The overall accuracy reached 97%.