Comparison of neural network and extended Kalman filter determination of kinematics from impact acceleration tests

This paper compares the performance of an artificial neural network (ANN) with that of an extended Kalman filter (EKF) for use in processing accelerometer data collected during impact acceleration tests. Both the EKF and the ANN are required to produce subject kinematics from the noisy readings. The ANN is much simpler to implement, and gives desirable results. Moreover, the ANN implementation, unlike the EKF implementation, does not require a model of the system