PHYSIOLOGICAL ASPECTS FOR SONOGRAMS BUILDING AND SPECTRUM RESTORE DISTORTED SPEECH VOCALIZATIONS
The aim of the study is to substantiate the physiological approach to spectral analysis-synthesis of the speech signal (RS), which involves taking into account the biometric parameters unique for each person due to the physical nature of his speech-forming tract, necessary for solving the problems of protection of speech information (SI) from falsification and forgery, as well as for restoration of speech intelligibility (RS) at distorted areas of voice interfaces. The main groups of physical parameters of the human articulatory system used for this purpose are highlighted. The influence of the time-varying shape of the envelope of the harmonic components of the RS on the obtained frequency resolution when performing a short-time Fourier transform (STFT or CPF) with a Gaussian weighting window is analyzed. A comparison of the resolution of the Fourier transform with a Gaussian window for spectral analysis-synthesis of sinusoidal signals with different envelope functions is presented. The possibility of application of the obtained results for reconstruction of the harmonic structure of distorted parts of speech vocalizations (SVs) is shown.
Authors
- Sergey V. Dvoryankin (ORCID: https://orcid.org/0000-0001-6908-0676)
- Nikita S. Dvoryankin (ORCID: https://orcid.org/0000-0002-1580-7179)
- Mikhail V. Alyushin (ORCID: https://orcid.org/0000-0001-7806-3739)
- Alexander M. Alyushin
Institutions
- Moscow State Linguistic University (RU)
- National Research Nuclear University MEPhI (RU)
Publication Details
- Journal
- Bezopasnost informacionnyh tehnology
- Published
- 2025-05-01
- DOI
- https://doi.org/10.26583/bit.2025.2.03
- Primary Topic
- Speech Recognition and Synthesis
- Type
- article
- Field-Weighted Citation Impact
- 0.00