Experimental Evaluation of MIMO-WLAN-based Object Detection with Reflectors

Various object detection schemes using channel state information (CSI) in wireless local area networks (WLANs) such as IEEE802.11ac have been investigated recently. For further detection performance improvement, adopting a proper feature selection technique and using an appropriate antenna placement to obtain more effective CSI is important As proposed herein, a device-free WLAN-based object detection scheme with reflectors is developed, where the feedback CSI in WLANs (i.e., beamforming weight information) is captured at an off-the-shelf WLAN device and is used as feature information for machine learning (ML)-based object detection. In this scheme, reflectors are placed to create a multipath rich condition in an observation area. By experimentation, we demonstrate that reflectors can improve object detection performance. Additionally, we clarify that proper antenna placement effectively improves the detection accuracy, even when using a small number of antenna elements in an indoor environment. Experiment results demonstrate that the object detection performance of WLAN systems in an indoor environment can be improved by selecting antenna placement appropriately and by using measured concatenated CSI as effective feature information.