Ambient WiFi signals enable indoor activity recognition with language descriptions
CSI-TextBench: A Dataset and Benchmark for Language-Grounded Ambient Sensing Perception
Artificial IntelligenceEmerging Technologies
Summary
Understanding what happens indoors using WiFi signals could help in homes and smart buildings. The authors created a large dataset linking WiFi signals, which capture movement and environment, to written descriptions. This lets AI systems learn to recognize indoor scenes and activities using natural language. Their method matches or beats traditional classifiers and adapts to new tasks more easily. This work helps connect these signals to language-based AI, improving indoor awareness without cameras or microphones.
What this means in practice
- •For smart building designers: Use WiFi signal data paired with natural language to enable context-aware building automation without additional sensors.
- •For assistive technology developers: Develop assistive devices that interpret indoor activities through WiFi signals described in natural language to support occupants unobtrusively.
- •For security system integrators: Incorporate language-grounded ambient WiFi sensing into security setups to better detect and describe indoor human activity patterns.$Commercial implications: Enables new products that use WiFi-based activity recognition for enhanced security monitoring and reporting.
Authors
Guozhen Zhu, Yuqian Hu, Sakila S Jayaweera, Wei-Hsiang Wang, Beibei Wang, K. J. Ray Liu
Abstract
Indoor context deciphered by ambient sensing underpins many applications, from assisted living to smart buildings, yet remains weakly connected to the language-aligned paradigm of modern multimodal systems. Through billions of IoT devices, ambient WiFi Channel State Information (CSI) captures the indoor context of human activity and environmental dynamics passively via ubiquitous, always-on infrastructure. However, unlike vision or audio, CSI lacks naturally paired text like captions or transcripts, limiting its integration with language-driven AI systems. We present CSI-TextBench, a large-scale dataset and benchmark for aligning CSI data with natural language, unifying 3 sensing benchmarks into 285,790 samples across 7 tasks and pairing them with 2,291 hierarchically structured, label-derived descriptions via many-to-many assignment to construct 2.0 million CSI-text training tuples. Benchmarking 5 alignment methods from 4 paradigm families, we show that contrastive CSI-text alignment matches supervised classifiers without class labels, adapts to unseen classes with few samples, and responds to open-vocabulary natural-language prompts. A single unified model supports compositional scene understanding, answering joint queries over activity, proximity, and identity, achieving 86.9% top-1 accuracy on 140 composite scenes, outperforming task-specific classifier ensembles. CSI-TextBench enables ubiquitous ambient sensing as a language-indexed perception layer, enabling indoor contextual awareness in language-driven AI systems.