IMPROVEMENT OF THE ALGORITHM FOR RESTORING THE LINGUISTIC COMPONENT OF SPEECH INFORMATION USING THE PHONEME–FORMANT METHOD FOR ASSESSING ITS SECURITY LEVEL

Abstract

The article presents an improved algorithm for restoring the linguistic component of speech information using the phoneme–formant method in tasks of assessing its protection level. A comprehensive analysis of the historical development of articulation and formant approaches is conducted — from the classical studies of Bell Laboratories and Harvey Fletcher to modern digital techniques of speech intelligibility evaluation established in ANSI S3.5-1997 and ANSI S3.5-2007 standards. It is shown that the evolution of speech-signal analysis methods has directly influenced the formation of information protection systems, particularly in monitoring channels of acoustic leakage. A ten-step algorithm for speech signal analysis and reconstruction is proposed, integrating spectral, statistical, and contextual approaches. The algorithm includes speech activity detection (VAD), signal segmentation, formant extraction using Fourier and linear predictive analysis (FFT, LPC), and construction of a probabilistic triphone model. The consideration of phase shifts, coarticulation effects, and inter-formant intervals ensures high reconstruction accuracy and robustness even under low signal-to-noise ratio (SNR) conditions. Special attention is given to analyzing the influence of spectral distortions of formants on phoneme recognition and determining statistical regularities of diphone and triphone occurrence in Ukrainian speech. This enabled the creation of an adaptive model combining acoustic and linguistic features for more reliable restoration of the linguistic component of a message. The developed algorithm can be applied in technical information protection practice for objective evaluation of speech-channel security and for recovering damaged or noisy recordings. Its structure fully complies with Ukrainian national standards ND TZI 1.6-005-2013 and ND TZI 3.7-003-2023, which regulate the methodology for analyzing speech-information leakage channels and assessing protection efficiency. The obtained results have dual practical significance: they can serve as a foundation for expert systems of speech information leakage control and for systems of automatic speech recognition, synthesis, and restoration requiring high phonemic precision. The proposed phoneme–formant approach demonstrates universality and the ability to integrate with modern digital audio-analysis tools, thus improving both information security and the performance of acoustic control systems.