METHODS AND ALGORITHMS FOR SPEECH SYNTHESIS BASED ON TEXT

The article is devoted to the development of a computer model based on methods and algorithms for speech synthesis, covering the basis for using the created speech corpus of the Tajik language. The task of speech processing was determined based on its digital image. Methods of speech synthesis are discussed, such as parametric speech synthesis, concatenation speech synthesis, and complete synthesis of speech according to rules. Based on the results obtained, the speech synthesis algorithm was translated into the concatenation method and proposed as a flowchart. Problems of selecting and linking parts of speech are proposed, taking into account changes in the values of synthetic speech measurements to the type of mathematical models. To solve the problem of speech synthesis, we found a solution: preliminary text analysis; selection and connection of speech components of natural language from a database based on an automatic decomposition algorithm; changing the values of phonetic and prosodic sizes of synthetic speech using a sinusoidal speech synthesis model; developing of computer programs to ensure sound synthesis in Tajik language. The results under consideration were obtained at the Khujand City Scientific Center NAST within the framework of the budget project «Development of Tajik language speech corpus for solving problems of computational linguistics», approved under number 0123TJ1547.

computational linguistics, Tajik language, computer model, speech synthesis, speech processing technologies, speech corpus.

1. Dvoryankin S. V., Dvoryankin N. S., Alyushin A. M. Rapid synthesis of audio signals from spectrogram images in the tasks of speech information protection // Issues of Cybersecurity. – 2024. – No. 5 (63). – pp. 34–46.

2. Khudoyberdiev Kh. A. Modeling of an automatic text processing system in the Tajik language // International Journal of Open Information Technologies. – 2023. – Vol. 11, No. 3. – pp. 27–33. – EDN KRBOBH.

3. Khudoyberdiev Kh. A. On the Tajik text-to-speech synthesizer // New Information Technologies in Automated Systems. – 2013. – No. 16. – pp. 273–276. – EDN RPDEPF.

4. Khudoyberdiev Kh. A. The Algorithms of Tajik Speech Synthesis by Syllable // ITM Web of Conferences. – 2020. – Vol. 35. – p. 07003. – DOI: 10.1051/itmconf/20203507003. – EDN MLGZSU.

5. Khudoyberdiev Kh. A., Muzaffarov D. Z., Ashurova Sh. N. Development of the Tajik speech corpus for solving some tasks of computational linguistics // Bulletin of PITTT named after academician M. S. Osimi. – 2023. – No. 2(27). – pp. 7–14. – EDN FMCKBZ.

6. Lobanov B. M. An algorithm for text segmentation into syntactic syntagmas for speech synthesis [Text] / Lobanov B. M. // Proceedings of the International Conference “Computational Linguistics and Intellectual Technologies” (Dialogue’2008). – Moscow: Nauka, 2008. – pp. 323–529.

7. Maksudov A. T., Khudoyberdiev Kh. A., Ashurova Sh. N. On the issue of creating a corpus of Tajik speech // University Bulletin. Series of Natural and Economic Sciences. – 2024. – No. 2(69). – pp. 9–15. – EDN RNDYRX.

8. Maksudov A. T., Khudoyberdiev Kh. A., Solieva M. T. On the system of automatic recognition of key words in conversational speech // Polytechnic Bulletin. Section: Intellect, Innovation, Investment. – 2024. – No. 2(66). – pp. 57–60. – EDN SYWKAD.

9. Nikonorov S. A., Bogolyubov A. N. Wavelet analysis of audio signals and speech synthesis // Scientific Notes of the Faculty of Physics of Moscow University. – 2018. – No. 6. – pp. 1860601-1.

10. Tsirulnik L. I., Zhadinets D. V., Lobanov B. M., Sizonov O. G. Algorithms for synthesis of prosodic speech characteristics from text in the “Multiphon” system [Text] // Computational Linguistics and Intellectual Technologies: Proceedings of the International Conference Dialogue’2007, Bekasovo, May 30 – June 3, 2007. – Moscow: RGGU Publishing Center, 2007. – pp. 550–558.

2026-03-31