Koksal Nasrettin, Ghannoum AbdulRahman, Melek William, Nieva Patricia
Mechanical and Mechatronics Engineering Department, University of Waterloo, Waterloo, ON N2L 3G1, Canada.
Department of Chemical Engineering, Qatar University, Doha P.O. Box 2713, Qatar.
Sensors (Basel). 2025 Apr 22;25(9):2638. doi: 10.3390/s25092638.
Occupancy monitoring (OM) and the localization of individuals within indoor environments using wearable devices offer a very promising data communication solution in applications such as home automation, smart office management, outbreak monitoring, and emergency operating plans. OM is challenging when developing solutions that focus on reduced power consumption and cost. Bluetooth low energy (BLE) technology is energy- and cost-efficient compared to other technologies. Integrating BLE Received Signal Strength Indicator (RSSI) signals with machine learning (ML) introduces a new Artificial Intelligence- (AI-) enhanced OM approach. In this paper, we propose an Intelligent Bluetooth Virtual Door (IBVD) OM system for the indoor/outdoor tracking of individuals using the interaction between a BLE device worn by the occupant and two BLE beacons located at the entrance/exit points of a doorway. ML algorithms are used to perform intelligent OM through pattern detection from the BLE RSSI signal(s). This approach differs from other technologies in that it does not require any floorplan information. The developed OM system achieves a range between 96.6% and 97.3% classification accuracy for all tested ML models, where the error translates to a minor delay in the time in which an individual's location is classified, introducing a highly reliable indoor/outdoor tracking system.
占用监测（OM）以及使用可穿戴设备在室内环境中对人员进行定位，在诸如家庭自动化、智能办公管理、疫情监测和应急操作计划等应用中提供了一种非常有前景的数据通信解决方案。在开发注重降低功耗和成本的解决方案时，占用监测具有挑战性。与其他技术相比，蓝牙低功耗（BLE）技术具有能源和成本效益。将BLE接收信号强度指示（RSSI）信号与机器学习（ML）相结合，引入了一种新的人工智能（AI）增强型占用监测方法。在本文中，我们提出了一种智能蓝牙虚拟门（IBVD）占用监测系统，用于通过占用者佩戴的BLE设备与位于门口入口/出口点的两个BLE信标之间的交互来对人员进行室内/室外跟踪。ML算法用于通过从BLE RSSI信号中进行模式检测来执行智能占用监测。这种方法与其他技术的不同之处在于它不需要任何平面图信息。所开发的占用监测系统对于所有测试的ML模型实现了96.6%至97.3%的分类准确率范围，其中误差转化为对人员位置进行分类时的轻微延迟，从而引入了一个高度可靠的室内/室外跟踪系统。
