WiFi CSI-Based Device-free Multi-room Presence Detection using Conditional Recurrent Network

Abstract

Human presence detection via camera-based monitoring systems has been well-adopted in various applications including smart homes, factories, and hospitals. However, its privacy concerns have been raised in many occasions such as daycare centers and homes with elderly living alone. In recent years, literatures adopting wireless signals were proposed to resolve privacy issues for presence detection; nevertheless, existing works can only be applied in a single room scenario. In this paper, we are the first work to propose a device-free multi-room human presence detection system based on efficient star network topology. Our proposed conditional recurrent architecture-based multi-room presence detection (C-MuRP) system extracts both spatial and temporal features from the Wi-Fi channel state information (CSI). Associated with a voting scheme, the proposed novel deep-learning architecture classifies the states of multi-room with the condition on current waveform to emphasize present feature states. Real-time experimental results showed that our proposed C-MuRP system can achieve higher accuracy for multi-room presence detection compared to existing methods.