Constraint-conscious smoothing framework for the recovery of 3D articulated motion from image sequences

Multi‐viewpoint gesture recognition by an integrated continuous state machine

“…On the other hand, the Kalman filter has been defined as a probabilistic process and its state estimation has been formalized by sequential maximization of the state likelihood. While the Kalman filter has a probabilistic basis and can represent the reliability of both predicted states and external inputs, it is mostly applied in tracking systems with prediction and smoothing [7,8]. We propose a continuous state machine that integrates sequential optimization by means of Kalman filters and class discrimination methods based on recurrent neural networks.…”

A Survey of Computer Vision-Based Human Motion Capture

“…Another promising approach is to use future measurements when processing the current data, i.e., allow a lag in the output. This helps to resolve ambiguities [16,127].…”

Pose Estimating the Human Arm Using Kinematics and the Sequential Monte Carlo Framework

“…Therefore kinematic constraints are often applied in order to prune the solution-space, e.g., the bending of the elbow is between 0 and 145 o . This may be used directly to partition the solution-space into legal and illegal regions, as in (Segawa et al, 2000), or the constraints may be defined as forces acting on an unconstrained state phase (Wren et al, 2000). The fact that two human body parts cannot pass through each other also introduces constraints.…”

“…By the latter is meant that the model guides the estimation process, as opposed to the former where the model is an integrated part of the estimation process. For example, in the work by (Segawa et al, 2000) a particular pose of an arm is estimated by filtering past and future image measurements of the arm. The filtering is done using an Extended Kalman Filter wherein the joint angle limits of the model are incorporated.…”