Speechlike signal synthesis module for information security systems

Abstract

A comparative analysis of the effectiveness of noise signals of different types in order to reduce the intelligibility of masked speech in the case of leakage through direct acoustic and vibro-acoustic channels. As a test signal used articulation phrase tables from the standard STB GOST R 50840-2000. A technique for testing the effectiveness of various types of noise signals is proposed. One of the results of the work is the development of a method for the formation of an effective masking signal for active protection of acoustic information, which is a correlated speech-like noise, formed directly from the masked speech signal, the level of which can be 7-9 dB below the white noise level to ensure the required value of speech intelligibility. Speech-like interference synthesis method includes the following stages: speech detection, speaker verification, speech segmentation, speech classification, speech-like interference synthesis. Synthesis of speech-like interference is based on compilation of allophones by phonemic text generated taking into account language statistics. It allows form speech-like interference in Russian, English or Arabic languages.