A low-pass filter is a signal processing technique that attenuates high-frequency components in a signal while preserving lower-frequency components below a defined cutoff threshold. In WiFi/CSI sensing, it is applied to raw CSI amplitude or phase time series to suppress high-frequency noise, hardware-induced fluctuations, and environmental interference that would otherwise obscure the subtle, slowly varying patterns associated with human activities such as gait cycles or respiration. Common variants used in the field include the Butterworth filter, moving average filter, and Savitzky-Golay filter, each offering different trade-offs between sharpness of frequency cutoff, phase distortion, and computational cost, with the choice of variant and cutoff frequency being critical to preserving the physiologically or behaviorally relevant signal content.
Dictionary term