UDC 004.522
AUTOMATIC IDENTIFICATION OF A PERSON ACCORDING TO VOICE USING THE ALGORITHM BASED ON GAUSSIAN MIXES MODEL
D. I. Lukyanov, associate professor, department of radiocontrol and communication, RSREU, PhD; This email address is being protected from spambots. You need JavaScript enabled to view it.
A. S. Mikhailova, master student, RSREU; This email address is being protected from spambots. You need JavaScript enabled to view it.
Methods of automatic identification of a person according to voice are researched. The aim of the work is determination of the most effective algorithm of identification of a person on the basis of speech signal providing high probability of correct identification. Pilot studies of the algorithm on the basis of method of mean square deviation, minimum distance, Gaussian mixes are conducted. Identification of a person using the algorithm on the basis of method of mean square deviation isn't completely invariant to initial RS as is effective for short speech messages containing one word. Identification of a person using the algorithm on the basis of method of minimum distance allows to carry out identification at the level of man/woman. One of the most effective methods is the method based on Gaussian mixes model. The optimum number of clusters for recognition is equal to 5. Implementation of the algorithm on the basis of model of Gaussian mixes allows to make correct identification with probability to 75 %.
Key words: recognitions of voice information, recognition systems, identification of a person according to voice, method of mean squared deviation, method of minimum distance, method swept-cepstral conversions of RS range, cepstrum, model of Gaussian compounds, optimum number of clusters.