It is claimed that the spectral analysis schemes, which are based on an explicit model of the speech signal (homomorphic and linear predictive analysis), are sensitive to errors in the model. In particular, speech is usually recorded in the presence of some type of background noise. In this research paper, the changes in the spectral envelope caused by the presence of additive white noise will be explored.

Keywords : Fourier analysis, Homomorphic analysis, Linear Prediction Coding