Singular Value Decomposition (SVD) is a matrix factorization technique that decomposes a complex-valued CSI matrix into three matrices — typically denoted U, Σ, and V* — separating it into orthogonal signal components ranked by their energy contribution. In WiFi sensing, SVD is used to extract dominant propagation paths or isolate individual subjects' channel signatures from composite multi-antenna or multi-subcarrier CSI measurements, enabling spatial separation of overlapping signals such as those produced by multiple people in a near-field environment. Key variants relevant to the field include truncated SVD, which retains only the top singular values to suppress noise, and the use of the right singular vectors (V matrix) to characterize per-user channel subspaces, as leveraged in systems like MUSE-Fi for physically separating the contributions of multiple persons from a shared wireless channel.
Source Papers
- A Robust CSI-Based Scatterer Geometric Reconstruction Method for 6G ISAC System ↗ — A Robust CSI-Based Scatterer Geometric Reconstruction Method
- Channel State Information from Pure Communication to Sense and Track Human Motion: A Survey ↗ — Channel State Information from Pure Communication to Sense a
- Efficient machine learning for Wi-Fi CSI-based human activity recognition using fast Monte Carlo based feature extraction ↗ — Efficient machine learning for Wi-Fi CSI-based human activit
- MUSE-Fi: Contactless MUti-person SEnsing Exploiting Near-field Wi-Fi Channel Variation ↗ — MUSE-Fi: Contactless MUti-person SEnsing Exploiting Near-fie
- Tool release ↗ — Tool release