THE SYSTEM OF RECOGNITION OF KEY PHRASES IN SPEECH ON THE EXAMPLE OF TAJIK LANGUAGE

This article presents research analyzing the results obtained using modern dynamic programming algorithms for keyword and phrase recognition. A Hidden Markov Model (HMM) is employed for phoneme modeling, a crucial component in the keyword recognition process. This HMM allows for consideration of the probabilistic characteristics of phonemes, ultimately resulting in high accuracy of keyword phrase recognition by refining sounds and words. Keyword phrases are represented as a sequence of sound elements in the form of syllable transcriptions. The results of the keyword phrase search are presented on a relatively small dataset of participant voice recordings. A specialized algorithm for searching keyword phrases within sequences of speech phonemes, represented as syllables, has been developed. In the context of developing Tajik language speech corpus, the proposed algorithm is intended for use in searching for phonetic features within large volumes of speech data. The outcome of the article presents a system for recognizing phrases and key words in speech using the example of Tajik language.

speech recognition, speech phonemes, speech corpus, hidden Markov model, search for keyphrases in speech, Tajik language.

1. Dadoboev, A. A. / Mathematical models of voice recognition. // Payomi Politekhniki (Polytechnic Bulletin): Intellect, Innovation, Investmen. – 2023. – (1) 61, 27-33. – EDN VNMJGH.

2. Dadoboev, A. A. / Speech recognition algorithm. // Nomai Donishgoh (University Letter): Series of Natural and Economic Sciences. – 2024. – (3) 67, 12-15.

3. Khudoiberdiev, Kh. A. / Segmentation of a speech signal based on syllabic structures of Tajik language. // New Information Technologies in Automated System. – 2018. – 21, 181-182. – EDN XNRSBF.

4. Khudoiberdiev, Kh. A., & Ashurzada, B. Kh. / Syllabification of the speech recognition process based on Tajik language speech. // Politekhnicheskiy Vestnik (Polytechnic Bulletin): Series Intellect, Innovation, Investment. – 2022. – 2 (58), 39-42.- EDN VNMJGH.

5. Maksudov, A. T., Khudoiberdiev, Kh. A., & Solieva, M. T. / On the system of automatic keyword recognition in conversational speech. // Payomi Politekhniki (Polytechnic Bulletin): Intellect, Innovation, Investment. – 2024. – 2 (66), 57-60. – EDN SYWKAD.

6. Vishnyakova, O. A., & Lavrov, D. N. / An algorithm for phoneme segmentation based on the analysis of the rate of change of energy of discrete wavelet transform. // Informational Technologies. – 2011. – 4, 146–152.

2026-03-26