A median filter is a nonlinear signal processing technique that replaces each data point in a time series with the median value of a surrounding window of samples, effectively removing impulsive noise and short-duration outliers while preserving the underlying signal structure. In WiFi/CSI sensing, it is commonly applied as a preprocessing step to clean raw CSI amplitude or phase streams of transient spikes caused by multipath interference, hardware noise, or environmental disturbances, thereby improving the reliability of subsequent feature extraction and activity recognition. Key variants include sliding window median filters of varying window lengths, where a larger window provides stronger noise suppression at the cost of temporal resolution, and frequency-selective applications where the filter is applied independently to each CSI subcarrier to handle per-subcarrier noise characteristics.
Dictionary term