CSI Sampling for Room-by-Room Device Grouping in Practical Environments

To reduce the cost of Internet of Things (IoT) system deployment, we are focusing on the cost reduction of device location information setup and developing a room-by-room device grouping system. In our previous work, we presented a room-by-room IoT device grouping based on wireless LAN (WLAN) channel state information (CSI) using unsupervised learning. The performance in a practical environment, however, is significantly degraded due to the nonuniform time distributions of where people stay in each room. In this paper, we present CSI sampling, namely, a CSI data selection method, relying on independent component analysis (ICA) to improve the device grouping performance in a practical environment. An experimental evaluation conducted in a practical environment reveals that our CSI sampling greatly improved device grouping performance with an adjusted Rand index (ARI) of up to 44.9%.