On the prospects for the introduction of biometric information into a graphical speech signal format

Savelyev I., Antipenko A.

Currently, due to the improvement of machine learning algorithms, as well as a significant reduction in the cost of computing power of server equipment, deepfake audio technologies are becoming more and more popular, which allow artificially synthesizing (faking) the speaker's voice. The article discusses the main areas of application of deepfake audio, the threats that these technologies carry, as well as ways to stop them, one of which is the use of a speech signature. A software and hardware complex are presented that allows, based on the method of calculating the phase characteristics of voice vocalisms, including embedding biometric information in the graphic format of a speech signal.

Keywords:voice vocalisms, speech information protection, intelligibility, speech signal, phase characteristics, sinusoidal model, biometrics.