Multirate Kalman Filter Rejects Impulse Noise in Frequency-Domain-Multiplexed Tracker Measurements.

Abstract

Frequency domain multiplexing (FDM) is a useful for making multiple measurements simultaneously, such as in optical and electromagnetic position trackers. Much interference is periodic (e.g., AC power harmonics), and is rejected well by FDM, but impulse disturbances are also common. Impulses corrupt the entire spectrum for a short period, and are better rejected in the time domain. Nonlinear blanking is a simple way to suppress impulses, but cannot be easily realized when the required dynamic range is large, and problematic noise may be far smaller than the signal. The described multi-rate Kalman filter upsamples the prediction to the input rate so that impulse departures from the predicted signal are easily detected and blanked out. Also, noise levels in unused adjacent channels are used to estimate measurement noise so that the Kalman filter adapts more slowly when noise is high, keeping output noise roughly constant even in the presence of longer noise bursts.