The problem of personality recognition by voice using adaptive Kalman filter is considered. The extracted features of the acoustic signal are used as features of biometric authentication person. A comparative table of speaker separation errors and evaluation speaker separation system using Kalman filter is presented.

Keywords: biometric authentication, voice, neural network compilation, adaptive Kalman filter

The result of the research is a method of complex biometric authentication. The method is implemented in the form of a software complex consisting of a subsystem biometric authentication by face image and a subsystem biometric authentication by voice. The training sample consisting of stored files biometric images (facial images and audio recordings) allows to reduce the error rates of the first and second kind in user recognition. The proposed method of biometric authentication is designed to improve the efficiency user recognition processes.

Keywords: authentication, biometrics, neural network architecture, training sampling