Sequential Importance Resampling (SIR) is a particle filter-based Bayesian inference method that iteratively propagates a set of weighted samples (particles) through a state-space model, updating their importance weights based on incoming observations and periodically resampling to concentrate computational effort on high-probability regions of the state space. In WiFi/CSI sensing contexts, SIR is particularly valuable for tracking occupancy counts and pedestrian positions by fusing noisy CSI measurements with probabilistic crowd or agent models, enabling robust real-time state estimation without requiring explicit device association. Key variants include the bootstrap particle filter, which draws proposal samples directly from the prior transition model for simplicity, and auxiliary particle filters, which incorporate future observation likelihood into the resampling step to improve efficiency in high-dimensional or low-noise sensing scenarios.
Source Papers
- Data Assimilation for Agent-Based Models ↗ — Data Assimilation for Agent-Based Models
- Device-free occupancy detection and crowd counting in smart buildings with WiFi-enabled IoT ↗ — Device-free occupancy detection and crowd counting in smart
- Recent trends in crowd analysis: A review ↗ — Recent trends in crowd analysis: A review