Statically Fused Converted Measurement Kalman Filters

Abstract

This chapter presents a state estimation method without using of nonlinear recursive filters when Doppler measurement is incorporated into the tracking system. The commonly used motions, such as the constant velocity (CV), constant acceleration (CA), and constant turn (CT), are represented in a pseudo-state space, defined from the product of target true range and range rate, by linear pseudo-state equations. Then the linear converted Doppler measurement Kalman filter (CDMKF) is presented to extract pseudo-state from the converted Doppler measurement, constructed by the product of the range and Doppler measurements. The output of the CDMKF is fused statically with that of the converted position measurement (range and one or two angles) Kalman filter (CPMKF) to produce target Cartesian state estimates. The accuracy and consistence of the estimator can be both guaranteed, since linear Kalman filters are both used to extract information from position and Doppler measurements.

Human Motion Enhancement via Tobit Kalman Filter-Assisted Autoencoder

“…Likewise, the observed measurements in human motion enhancement can be defined as the coordinates of skeletal joints, q , which correspond to the joint positions of the observed D-Mocap data. Given what we know of Newtonian physics, we can predict the state of the system using ( 4 ) and a constant velocity (CV) model [ 60 ], where Δ t is the period of the sampling of human motion data. The original D-Mocap data are represented in the form of a dynamic skeleton that has a fixed number of moving joints.…”

“…which correspond to the joint positions of the observed D-Mocap data. Given what we know of Newtonian physics, we can predict the state of the system using (4) and a constant velocity (CV) model [60],…”