Dynamic Time Warping (DTW) is a similarity measurement algorithm that computes the optimal alignment between two time-series sequences by allowing elastic stretching and compression along the time axis, enabling meaningful comparison of signals that may vary in speed, duration, or timing. In CSI-based human behavior recognition, DTW is particularly valuable because human activities such as walking, gestures, or respiration rarely unfold with perfectly consistent timing, and DTW allows matching of CSI signal patterns despite these natural temporal variations. Common variants used in the field include constrained DTW, which limits the warping path within a defined window (e.g., the Sakoe-Chiba band) to reduce computational cost and prevent pathological alignments, and weighted DTW, which assigns greater importance to certain segments of the time series during comparison.
Source Papers
- A Survey on Human Behavior Recognition Using Channel State Information ↗ — A Survey on Human Behavior Recognition Using Channel State I
- Channel State Information from Pure Communication to Sense and Track Human Motion: A Survey ↗ — Channel State Information from Pure Communication to Sense a
- CrossSense: Towards Cross-Site and Large-Scale WiFi Sensing ↗ — CrossSense: Towards Cross-Site and Large-Scale WiFi Sensing
- WiFi Sensing with Channel State Information ↗ — WiFi Sensing with Channel State Information