This study investigates the effectiveness of the Whisper model for audio-to-text transcription, specifically targeting the enhancement of accessibility for individuals with hearing impairments. The research focuses on the processing of audio recordings obtained from WhatsApp messenger, which often contain significant background noise that complicates speech recognition. To address this issue, advanced audio processing techniques were employed, including the use of the Librosa library and the Noisereduce package for noise reduction. The spectral gating methods applied in this study effectively diminished wind noise and other ambient sounds, allowing for clearer recognition of spoken content. To ensure the quality of the processed audio, we assessed its clarity using a SimpleRNN model. The training results demonstrated a progressive reduction in loss values across epochs, confirming the successful enhancement of audio quality. Once the audio files were adequately cleaned, we utilized the Whisper model, a sophisticated machine learning tool for speech recognition developed by OpenAI, to transcribe the audio into text. The transcription process yielded accurate Kazakh language output, despite the initial challenges posed by background noise. These findings underscore the critical role of high-quality audio input in achieving reliable transcription results and highlight the potential of machine learning technologies in improving communication access for hearing-impaired individuals. This study concludes with recommendations for future research, including the exploration of additional noise reduction techniques and the application of the Whisper model across various languages and dialects. Such advancements could significantly contribute to creating more inclusive digital environments and enhancing the overall user experience for individuals with hearing impairments.

This study investigates the effectiveness of machine learning methods in identifying earthquake-prone areas in Kazakhstan and its neighboring regions. By leveraging a comprehensive dataset encompassing significant earthquake data from 1900 to 2023, various machine learning algorithms were employed, including RandomForest, GradientBoosting, Logistic Regression, Support Vector Classification (SVC), K-Nearest Neighbors (KNeighbors), Decision Tree, XGBoost, LightGBM, AdaBoost, and MLPClassifier. The primary objective was to analyze and compare the performance of these models in predicting earthquake magnitudes and frequencies. The results reveal that certain algorithms significantly outperformed others in terms of accuracy, underscoring the potential of machine learning techniques to enhance earthquake prediction capabilities. Notably, XGBoost and RandomForest demonstrated the highest predictive accuracy, suggesting their suitability for application in seismic risk assessment. These findings offer valuable insights for governmental agencies engaged in disaster management and prevention planning, highlighting the practical implications of integrating advanced analytical techniques in their strategies. In addition to model performance analysis, a visual heatmap was generated to illustrate the geographical distribution of earthquake occurrences across the studied regions. This visual representation effectively identifies high-risk areas, serving as a crucial tool for local authorities and researchers in making informed decisions regarding safety measures and emergency preparedness. This research contributes to the expanding body of knowledge on earthquake prediction utilizing machine learning, emphasizing the necessity for continuous improvement in predictive models by incorporating additional environmental and geological factors. The implications of these findings extend beyond academic discourse, holding significant potential for enhancing public safety in regions vulnerable to seismic activity. As such, this study advocates for the integration of machine learning methodologies in disaster management frameworks to mitigate risks and enhance preparedness in earthquake-prone regions.

Қолданылған әдебиеттер Child Mind Institute. (2025). Speech and language disorders. https://childmind.org/topics/speech-and-language-disorders/ Diepeveen, S., et al. (2022). Process-oriented profiling of speech sound disorders. Children, 9(10), 1502. https://doi.org/10.3390/children9101502 Eadie, P., et al. (2024). Developmental speech and language disorders: A dataset for phonological processing research. Journal of Speech, Language, and Hearing Research, 67(2), 512–526. https://doi.org/10.1044/2023_JSLHR-23-00412 Hitchcock, E. R., et al. (2022). Acoustic characteristics of children’s speech. Journal of Speech, Language, and Hearing Research, 65(3), 987–1002. https://doi.org/10.1044/2021_JSLHR-21-00345 McLeod, S., & Crowe, K. (2020). Children’s consonant acquisition in 27 languages. American Journal of Speech-Language Pathology, 29(3), 1546–1571. https://doi.org/10.1044/2020_AJSLP-19-00168 Morgan, L., et al. (2021). Making the case for the collection of a minimal dataset for children with speech sound disorder. International Journal of Language & Communication Disorders, 56(5), 1097–1107. https://doi.org/10.1111/1460-6984.12650 Mussabekova, G., et al. (2021). KazakhTTS: An open-source dataset for Kazakh text-to-speech synthesis. arXiv preprint arXiv:2109.09267. https://doi.org/10.48550/arXiv.2109.09267 National Institute on Deafness and Other Communication Disorders. (2025). Quick statistics about voice, speech, language. https://www.nidcd.nih.gov/health/statistics/quick-statistics-voice-speech-language Skeat, J., et al. (2024). Challenges in collecting speech data from children in low-resource languages. PLOS ONE, 19(2), e0298456. https://doi.org/10.1371/journal.pone.0298456 Sugden, E., et al. (2021). Long-term outcomes of childhood speech sound disorders: A systematic review. International Journal of Language & Communication Disorders, 56(4), 737–753. https://doi.org/10.1111/1460-6984.12623 Torres, M., et al. (2020). Influence of cognitive-linguistic abilities on speech sound production. Journal of Communication Disorders, 84, 105976. https://doi.org/10.1016/j.jcomdis.2020.105976 Wren, Y., et al. (2023). Multilingual approaches to speech sound disorder research. Clinical Linguistics & Phonetics, 37(4–6), 456–472. https://doi.org/10.1080/02699206.2022.2072501 Авторлар туралы мәліметтер № Авторлар туралы мәлімет 1 А.Б. Абен – ғылым магистрі, Қожа Ахмет Ясауи атындағы Халықаралық қазақ-түрік университеті (Қазақстан, Түркістан қ.), e-mail: arypzhan.aben@ayu.edu.kz Дінмұхаммед Ғабдуллаұлы – Қожа Ахмет Ясауи атындағы Халықаралық қазақ-түрік университеті Инженерия факультетінің Компьютерлік инженерия кафедрасының магистранты (Қазақстан, Түркістан қ.), e-mail: gabdulladimas@gmail.com 2 А.Б. Абен – магистр наук, Международный казахско-турецкий университет имени Ходжи Ахмеда Ясави (Казахстан, г. Туркестан), e-mail: arypzhan.aben@ayu.edu.kz Динмухаммед Габдуллаулы – магистрант кафедры компьютерной инженерии факультета инженерии Международного казахско-турецкого университета имени Ходжи Ахмеда Ясави (Казахстан, г. Туркестан), e-mail: gabdulladimas@gmail.com 3 A.B. Aben – Master of Science, Khoja Akhmet Yassawi International Kazakh-Turkish University (Turkistan, Kazakhstan), e-mail: arypzhan.aben@ayu.edu.kz Dinmukhammed Gabdullauly – Master’s student, Department of Computer Engineering, Faculty of Engineering, Khoja Akhmet Yassawi International Kazakh-Turkish University (Turkistan, Kazakhstan), e-mail: gabdulladimas@gmail.com 4 A.B. Aben – bilim yüksek lisansı mezunu, Hoca Ahmet Yesevi Uluslararası Kazak-Türk Üniversitesi (Kazakistan, Türkistan), e-posta: arypzhan.aben@ayu.edu.kz Dinmukhammed Gabdullauly – Hoca Ahmet Yesevi Uluslararası Kazak-Türk Üniversitesi Mühendislik Fakültesi Bilgisayar Mühendisliği Bölümü yüksek lisans öğrencisi (Kazakistan, Türkistan), e-posta: gabdulladimas@gmail.com

In this paper, we introduce a time-aware deep learning model designed to identify and predict signs of depression and suicidal ideation across social networks. Standard static text classifiers typically analyze updates in isolation; however, our method tracks the long-term progression of a person's emotional state by merging contextual language embeddings with temporal encodings and specific psycholinguistic markers. We gathered our primary dataset from Twitter, Reddit, and Facebook, ensuring all user histories were strictly anonymized and organized chronologically. The study evaluates multiple neural network architectures, specifically Temporal BERT (Bidirectional Encoder Representations from Transformers), time-encoded BiLSTM (Bidirectional Long Short-Term Memory), and a temporal transformer utilizing positional features. Our experiments demonstrate that factoring in the chronological dimension substantially boosts classification accuracy, allowing for the earlier detection of declining mental health. The Temporal BERT model achieved the highest F1 score (harmonic mean of precision and recall) and AUC (Area Under the Receiver Operating Characteristic Curve) values on several datasets, outperforming both standard (static) BERT and basic recurrent models. Analysis of temporal trajectories also allowed us to identify clear clusters of user behavior: stable, improving, and deteriorating - this makes conclusions more interpretable and helps us understand personal emotional dynamics. The early-warning module was evaluated at 7-, 14-, and 21-day prediction horizons and showed that risk-related deterioration patterns could be identified in advance of the reference event. Across all evaluated horizons, Temporal BERT demonstrated the strongest Recall@k performance, meaning that it more consistently captured at-risk users among the top-ranked predictions. This article emphasizes that depressive and suicide-related risk signals are often not evident in isolated posts but emerge through longitudinal behavioral patterns. The proposed approach may support earlier and more sensitive identification of elevated risk patterns in digital mental health monitoring settings. At the same time, such use requires strict ethical safeguards, rigorous anonymization, and human-in-the-loop oversight. Future research should extend the framework toward multimodal, multilingual, and socially contextualized modeling.

Real-time human movement analysis has become an essential component of modern computer vision applications, including sports performance assessment, rehabilitation monitoring, intelligent surveillance, and human–computer interaction. Although deep learning–based pose estimation methods provide accurate human body landmark detection, executing person detection on every video frame introduces considerable computational overhead, limiting their applicability on resource-constrained CPU-based systems. This paper proposes an efficient YOLO–ROI framework for real-time human movement analysis that combines scheduled YOLO person detection, region-of-interest (ROI) generation, MediaPipe Pose estimation, and lightweight landmark smoothing. Instead of performing object detection on every frame, the proposed framework periodically executes the YOLO detector while reusing the most recently detected ROI between detection cycles. Consequently, MediaPipe Pose continuously estimates body landmarks within the ROI, significantly reducing redundant object detection operations and improving computational efficiency. The proposed framework was evaluated using a self-collected dataset consisting of 200 videos recorded from 10 participants performing five physical exercises under controlled indoor conditions. Computational performance was analyzed under different ROI sizes, detection intervals, and MediaPipe Pose complexity levels using CPU-based execution. Experimental results demonstrate that the proposed optimization strategy achieves a maximum processing speed of 61.32 FPS with an average processing latency of 16.31 ms, enabling stable real-time operation without GPU acceleration. The obtained results further show that scheduled YOLO detection, ROI reuse, and lightweight landmark smoothing effectively reduce computational workload while preserving continuous pose estimation throughout video processing. The proposed framework provides a simple, computationally efficient, and easily deployable solution for real-time human movement analysis and can be readily integrated into sports analytics, rehabilitation systems, intelligent surveillance applications, and edge-based computer vision platforms.

This study is designed to investigate effective methods for artificial intelligence-based detection of hyperkinetic speech disorders using audio signals. The main objective of the study is to compare several machine learning models and create a platform for automatic diagnosis of speech disorders using the Hyperkinetic Dysarthria Speech (HDS) dataset. The data was pre-processed and acoustic features (MFCC, Chroma, Spectral Centroid, Zero-Crossing Rate) were extracted. The models compared included Logistic Regression, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree, Random Forest, Gradient Boosting, AdaBoost, Extra Trees, XGBoost, LightGBM, CatBoost, and Multi-Layer Perceptron Classifier (MLPClassifier). According to the results, the Logistic Regression, KNN, SVM, CatBoost, and MLPClassifier models reached the level of 1.0 in terms of Accuracy, Precision, Recall, and F1-Score metrics, which indicates that pathological features in speech signals are classified with high accuracy. The research resulted in the development of the Streamlit platform based on the MLPClassifier model, which allows users to upload audio files or record speech in real time. This system reduces the subjectivity of traditional clinical assessments, facilitates early diagnosis, and helps improve the quality of life of patients. However, limitations of the study include the availability of data mainly in Russian and the relatively small size of the dataset (approximately 760 files), which limits the generalizability of the models to other languages. Also, the stability in noisy environments requires additional testing. Future studies are planned to incorporate multilingual data, use deep learning architectures, and validate the platform in clinical settings. This work will strengthen the role of computational hearing and machine learning in the detection of speech pathology and contribute to innovations in the field of neurological diagnostics and therapy.