Implementation of linear prediction techniques in state estimation

Abstract

Three different linear prediction coefficients (LPC) techniques are employed lo restore missing data in the process of state estimation. The conventional Normal Equation method has been found computationally expensive. Alternatively. Levinson Durbin Algorithm (LDA) considerably reduces this computational cost by avoiding the larger matrix inversions involved in the computation of LPC. However, LDA has been found suffering from a larger dynamic range in the values of LPC, An alternate method - Leroux Gueguen Algorithm (LGA) eliminates the problem associated with dynamic range in a stationary-point scenario by taking the application of Schwartz inequality in computation of this method. The main course of this work is to reduce the computational complexity of the Normal Equation when integrated with Kalman filter with that of LDA and LGA methods which do not require on matrix inversion in the computation of LPCs.