The article considers the problem of improving the methods of human identification using biometric features, in particular, the signals of an electroencephalogram. The authors present the results of scientific research into human identification by dynamics of frequency changes of the brain using the known architecture of artificial neural networks: AlexNet and Mobile Net 2. The basic hypothesis is formulated that the waves registered by sensors from head leads are unique for each person. The authors describe the preparation of experimental data on the basis of electroencephalogram signals received as a result of experiments on the formation of steady-state visual evoked potentials in a group of people with the subsequent creation of an applied database. The achievability of the set task was based on assessments of the relevance and representativeness of the obtained data. Frequency-time characteristics were taken as training datasets. Using deep machine learning technology, two classification models are obtained that allow identifying the identity of a person with a probability of 70%. The evaluation of the adequacy of the obtained classification models is carried out with PCA and t-SNE algorithms; the efficiency of their work is evaluated. The results of deep machine learning and machine classification tasks were confirmed by assessments of the adequacy of the obtained models. As a result, the authors confirm the main hypothesis.