The performance of the autocorrelation method of LP analysis has always been limited when the speech signal is corrupted by noise. In order to overcome these limitations a new approach for LP analysis needs to be designed.  A number of efficient LP analysis methods have been proposed. These are Moving maximum Method, Moving maximum Method and Average threshold Method. In this paper Conventional Auto correlation method is compared with the efficient LP analysis methods.

In this paper, the robustness of the proposed methods is measured between the spectrum of the efficient LP analysis method on clean speech and the spectrum of the efficient method on speech signal affected by noise. As the level of noise distorting the speech is increased, the LPC spectrum resulting from the proposed method is expected to keep its general shape.