In this paper, the text-independent method of person voice identification based on the features extraction from speech signal that characterize the linear prediction of the behavior of the autocorrelation function of the voice signal cepstrum are considered and developed.