Signal modeling and prediction using neural networks
SIGNAL MODELING AND PREDICTION USING NEURAL NETWORKS P. A. Ramamoorthy, G. Govind and V. K. Iyer. Mail Location #30, Department of Electrical and Comp...
SIGNAL MODELING AND PREDICTION USING NEURAL NETWORKS P. A. Ramamoorthy, G. Govind and V. K. Iyer. Mail Location #30, Department of Electrical and Computer Engineering, University of Cincinnati, Cincinnati, OH 45221 USA. Modeling (or system identification) and prediction problems are of great importance in the fields of signal processing and control systems. These enable the input/output relation of unknown systems to be modeled by a set of parameters. For stationary systems these parameters fully characterize the system and future outputs can be found using these parameters. In certain cases the plant could be noisy, in which case the solution models everything but the noise. Adaptive algorithms in use assume linearity of the unknown system and fit linear models to approximate the input/output relationships. Most of the systems in real-life are nonlinear but to minimize the complexity of analyzing these systems they are approximated as linear. Some work has been done in the area of nonlinear controls but is based on pure mathematics. Error propagation neural networks have been shown to learn the mapping of a static input pattern to a static output pattern. In this work, these inherently nonlinear neural networks are being explored to model and predict nonlinear processes. This could enable definition of a nonlinear model by a set of parameters. The neural network has been used for modeling autoregressive and moving-average processes and the results are comparable to those by linear techniques. Performance of the neural network is analyzed for nonlinear processes in comparison with linear adaptive techniques and simulation results presented.