Reproduce “CRONOS: Colorization and Contrastive Learning for Device-Free NLoS Human Presence Detection using Wi-Fi CSI” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Li-Hsiang Shen, Chia-Che Hsieh, An-Hung Hsiao, Kai-Ten Feng
A device-free human presence detection system, CRONOS, accurately identifies stationary individuals in non-line-of-sight environments by integrating dynamic recurrence plots and novel color-coded CSI ratio images with a three-stage supervised contrastive learning framework. It achieves an average F1-score of 95.28% in a conference room, demonstrating an 18.63% improvement for the NLoS static case compared to prior benchmarks.
Key insights
- Color-coding CSI ratio images, which combine amplitude and phase while mitigating random phase offsets, effectively encodes subtle positional information crucial for differentiating challenging NLoS static scenarios.
- Supervised contrastive learning, particularly with the novel consultation loss, enhances the discriminative power of feature representations, leading to better separation of highly similar static states.
- Dynamically switching between feature types (dynamic RPs and static CSI ratio images) based on a learned confidence for motion optimizes classification by leveraging each feature's strength where it is most effective.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.