To obtain access to full text of journal and articles you must register!
- Article name
- A technique for assessing the security of speech acoustic information from leakage through technical channels using convolutional neural networks
- Authors
- Volkov N. A., , volkovnikandr@gmail.com, Samara State Technical University, Samara, Russia
Ivanov A. V., , andrej.ivanov@corp.nstu.ru, Samara State Technical University, Samara, Russia
Karpova N. E., , nadevkar@mail.ru, Samara State Technical University, Samara, Russia
- Keywords
- deep neural networks / convolutional neural networks / signal-to-noise ratio / audio recording noise / speech intelligibility / spectrograms / low-frequency cepstral coefficients / assessment of the security of speech acoustic information
- Year
- 2025 Issue 3 Pages 22 - 32
- Code EDN
- AQPWAZ
- Code DOI
- 10.52190/2073-2600_2025_3_22
- Abstract
- The article considers a technique for assessing the security of speech acoustic information from leakage through technical channels using convolutional neural networks. The limitations of the generally accepted instrumental calculation technique based on the N.B. Pokrovsky formant method are analyzed. A formal estimation model is proposed based on the analysis of spectrograms and graphs of low-frequency cepstral coefficients, which were used for visual representation of acoustic signals. The training samples were formed taking into account differences in the speech spectra of the speakers and 16 gradations of verbal intelligibility determined by expert listening. Experiments have shown that the developed method provides a significantly lower discrepancy with the subjective assessment compared to the instrumental calculation method (the deviation does not exceed 0.05). The proposed technique takes into account conditions with a high level of masking interference, differences in the speech spectra of speakers, and also takes into account interference of the "speech chorus" type, which provides a more reliable assessment of speech security in real conditions.
- Text
- BUY for read the full text of article
- Buy
- 500.00 rub
