AMAR: Lightweight Attention-Based Multi-User Activity Recognition from Wi-Fi CSI
The paper introduces AMAR, a lightweight attention-based framework for multi-user activity recognition using Wi-Fi channel state information. It addresses challenges in recognizing overlapping activities from multiple users by employing a transformer-based architecture. The proposed system significantly improves activity prediction accuracy and reduces bandwidth requirements compared to existing methods.
- ▪AMAR formulates human activity recognition as a set prediction problem to handle multi-user scenarios.
- ▪The framework uses learnable query embeddings for simultaneous identification of multiple activities.
- ▪AMAR achieves an F1-score of 53.4%, outperforming the best benchmark by 7.8%.
- ▪The system reduces occupancy estimation error by 74% while minimizing bandwidth usage.
Opening excerpt (first ~120 words) tap to expand
Electrical Engineering and Systems Science > Signal Processing arXiv:2605.20649 (eess) [Submitted on 20 May 2026] Title:AMAR: Lightweight Attention-Based Multi-User Activity Recognition from Wi-Fi CSI Authors:Amirhossein Mohammadi, Hina Tabassum View a PDF of the paper titled AMAR: Lightweight Attention-Based Multi-User Activity Recognition from Wi-Fi CSI, by Amirhossein Mohammadi and Hina Tabassum View PDF HTML (experimental) Abstract:Wi-Fi-based human activity recognition (HAR) has emerged as a promising approach for contactless sensing, leveraging channel state information (CSI) collected from wireless transceivers.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
