A Method for the Real-Time Updating of Voice Samples in the Unified Biometric System

Abstract

We study the problem of automatic control of the quality of voice samples recorded and stored in the Unifi ed Biometric System. We propose a solution of the problem of timely updates of the collected samples required because their consumer qualities are lost in the course of time. A new indicator of the acoustic quality of voice samples in the Kullback-Leibler information metric is investigated and a new procedure is proposed for measuring this indicator at the times of contact of the users with the system with service requests. An example of practical implementation of the proposed method in the real-time mode is demonstrated. By using software specially developed by the authors, we performed a full-scale experiment and obtained quantitative estimates of the period of updating of voice samples. Some recommendations concerning their practical application are given. The accumulated results can be used for the purposes of development of new systems and technologies (and upgrading of the existing systems) of automatic quality control and adaptive renewal of the biometric personal data samples.