System Identification and Noise Cancellation: A Quantitative Comparative Study of Kalman Filtering and Neurai-Net Approaches

This paper reports on neural network approaches to system identification and noise cancellation tasks. Both linear and nonlinear systems in noisy environments can be handled without significant modification to the basic procedure. Results indicate that the neural network approach to system identification, and to noise cancellation problem is practicable, and has performance comparable to or superior to existing conventional algorithms.