Device-free human presence detection via infrared sensors or cameras has been well-developed in the past years. However, the infrared-based solutions suffer from misdetection problems with standstill people; while camera-based systems incur personal privacy issues. In recent years, wireless signals were adopted for presence detection, and channel state information (CSI) is one of the most popular information to achieve higher detection accuracy. Nonetheless, existing methods result in misclassification under non-line-of-sight (NLoS) static scenarios when the person stands still in the corner of the room. In this paper, based on multi-antenna Wi-Fi access points, we proposed the CSI ratio with coloring-assisted learning presence detection (CALPD) system that can detect human presence even when the person is motionless in the NLoS scenarios. The CSI ratio between antennas is illustrated on the complex plane to visualize the classification differences. Next, the RGB images are generated based on the proposed coloring-based classifier in order to distinguish and predict the final results. Field experimental results show that our proposed CALPD scheme outperforms other existing methods by achieving higher detection accuracy, especially under NLoS static scenarios.
CSI Ratio with Coloring-Assisted Learning for NLoS Motionless Human Presence Detection
2022-06-01
1773378 byte
Conference paper
Electronic Resource
English