Biometric features in the settings of the speech-likely signals for user authentication in voice driving systems

Dvoryankin S., Dvoryankin N.

Voice communications have been and remain one of the most common and convenient ways of human communication and remote human-machine interface. In recent years, there have been many reports of the successful creation and application of computer technology of speech cloning, with which you can successfully attack the system of speech control and automatic speech recognition, built-in information systems for various purposes. In this regard, the creation of methods for modeling speech-like signals with specified properties, used as specific audio markers, to confirm the authenticity of speech commands and messages is very important. Moreover, often in the process of remote interaction requires confirmation of the presence of a particular live subscriber with the verification of his authority, and not a robot acting in his place. The paper shows the possibility of using as identification audio markers, biometric characteristics of the legitimate user of protected resources, which in the form of graphic information (images) can be inserted into the spectral-time parameters of speech-like signals transmitted together with the passphrase (voice-key) via speeech communication channels during the negotiations of the user with the employee of the company, providing remote access to the protected resource.

Keywords:speech information protection, image analysis-synthesis, speech signal, binarization of spectrogram images, short-term Fourier transform.