The Recursive Estimator

In order to estimate the arm pose at each time step, we use a particular form of a recursive estimator, an Extended Kalman Filter. Such an estimator can be used based on the assumptions that:

In our particular case, the system whose state we want to estimate is a human arm, coupled with the camera system (the measurement device). Furthermore:

The original Kalman Filter theory assumes the system is linear, and that inaccuracies in the dynamical model and measurement noise can be modeled as gaussian processes. Since our system is highly nonlinear, an Extended Kalman Filter is used, where at each time step a linearization of the system about the current state prediction is computed. Furthermore, in order to speed up the estimator to run at real-time rates, a sequential measurement update scheme is used.