Zero-Shot Wi-Fi Human Activity Recognition via Signal-Language Alignment
July 30, 2026
Zero-Fi aligns Wi-Fi signal features with natural-language activity descriptions in a shared embedding space. This contrastive framework enables the recognition of unseen activity classes without requiring new labeled Wi-Fi samples or model adaptation.
HOW THIS AFFECTS YOU
●
researcherYou can extend Wi-Fi sensing capabilities to novel activities using semantic embeddings rather than supervised retraining.