In this paper, we propose a neural-network-based approach to acquiring angu
lar acceleration from a noisy velocity signal. Our scheme consists of two c
ascaded neural networks: neural network I (NN I) and neural network II (NN
II). NN I attenuates harmful measurement noise from the velocity input. NN
II further reduces the residual noise level, and gives the one-step-ahead p
rediction of the final acceleration signal. As an illustrative example, we
discuss the application of our method in the elevator velocity and accelera
tion acquisition problem. Two different kinds of neural network model are e
mployed here, the back-propagation neural network (BP) and the adaptive-net
work-based fuzzy inference system (ANFIS), to act as NN I and NN II. We als
o compare the performances of these two neural networks using numerical sim
ulations.