Comparing Autoregressive Moving Average (ARMA) Coefficients Determination using Artificial Neural Networks with Other Techniques

Abstract

Abstract—Autoregressive Moving average (ARMA) is a paramet-ric based method of signal representation. It is suitable for problemsin which the signal can be modeled by explicit known sourcefunctions with a few adjustable parameters. Various methods havebeen suggested for the coefficients determination among which areProny, Pade, Autocorrelation, Covariance and most recently, the useof Artificial Neural Network technique.In this paper, the method of using Artificial Neural network (ANN)technique is compared with some known and widely acceptabletechniques. The comparisons is entirely based on the value of thecoefficients obtained. Result obtained shows that the use of ANN alsogives accurate in computing the coefficients of an ARMA system.Keywords—Autoregressive Moving Average, Coefficients, BackPropagation, Model Parameters, Neural Network, Weight. I. I NTRODUCTION The use of modeling technique to predict or reconstruct adata sequence is concerned with the representation of data inan efficient technique [1]–[4], [6], [10], [13]. Signal modelinghave been used in radar application, geophysical application,Medical signal processing, ultrasonic tissue backscatter coeffi-cient estimation, speech processing, music understanding andmore recently in the field of Magnetic Resonance Imaging(MRI) reconstruction [2], [4], [6], [10], [13], [14], [16], [17].Signal modeling involves two steps steps [2], these are;1) Model selection: Choosing an appropriate parametricform for the model data2) Model Parameter determination: This include modelorder and model coefficients determination.Despite the success reported in the use of modeling tech-nique, two important problems constitutes challenges to theapplicability of this method, these are:1) Estimation of Model order: There have been variouseffort in determining a workable criteria for the determi-nation of an appropriate model order. The use of a modelwith an order too high over fits the data while the use ofa model with a low order leads to insensitivity to noise[2], [4], [6], [15].2) Estimation of model coefficient : The second impor-tant challenges mitigating against the use of modeling