Neural networks-based approach to the acquisition of acceleration from noisy velocity signal

In this paper, we propose a neural networks-based approach to acquire the angular acceleration from a noisy velocity signal. Our scheme consists of two cascaded neural networks: Neural Network I and II. Neural Network I attenuates the measurement noise from the velocity signal. Neural Network II further reduces the residual noise level, and calculates the final angular acceleration estimate. As an illustrative example, we discuss the application of our scheme in the elevator velocity and acceleration signal acquisition. Two different kinds of neural network models are employed: the back-propagation neural network (BP) and the adaptive-network-based fuzzy inference system (ANFIS). We compare the performances of these two neural networks by illustrative simulation experiments.