Abstract: In this study, we will present 2 psychoacoustic models used in an audio watermarking chain. The objective of this study is to present the psychoacoustics model of the human auditory mecanism and its implementation. The main application is the speech masking inherent in the human ear that will be used in the general outline of watermarking. The idea consists in modifying the power spectral concentration of the original code in order to insure a high sound audibility. It consists in generating a threshold starting from the masking phenomenon and a relative ponderation of the coefficients of MPA output signal contents.