Specific features of the formation of an acoustoelectric channel of speech information leakage

Abstract

We address the problem of assessing the protection level of the acoustoelectric channel of speech information with respect to leakages associated with specific operation principles of various electromechanical transducers. We analyze the scope of methodological and technical tools for the acoustical speech intelligence (ASI) with respect to corresponding technical channels of leakage of acoustic speech information (TCLSI). Special attention is paid to the regime of timed analysis of speech messages (distorted by noise and interference), allowing one to significantly improve the quality of the original audio signals. We list basic methods of noise reduction that can be used for processing the secondary signals of acoustoelectric leakage channels. We describe typical distortions that occur in the process of the acoustoelectric leakage channel formation. We examine the nature and the degree of the impact of various distortions on the measures of the speech information protection (SIP). We find the effect of the nonlinear distortions of the “upper limit” type (most typical for an acoustoelectric leakage channel) on the speech intelligibility to be in significant. Along with Pokrovsky’s static model of speech signals, we consider a dynamic model based on Pirogov’s phonetic function. The limitations of the static model are discussed, along with revealing the nature of the effect of the dynamic characteristics on speech intelligibility. We explain the effects of invariance with respect to linear distortions in the leakage channel. We perform an experimental study the results of which contradict, to a certain extent, the postulates of the formant theory used to assess the SIP level. We identify a number of mechanisms to improve the noise immunity of voice communications, allowing one to reconstruct speech signals (SS) distorted by noise and interference. We conclude with specifying a number of ways of improving SIP assessment methods in TCLSI.

Keywords

• voice acoustic reconnaissance
• acoustoelectric transducers
• formants
• phonemes
• information protection
• speech intelligibility
• speech signal