kaz29Вестник Казахстанско-Британского технического университетаHerald of the Kazakh-British Technical University1998-66882959-8109Казахстанско-Британский Технический Университет10.55452/1998-6688-2026-23-1-163-172kaz29-2510Research ArticleКОМПЬЮТЕРНЫЕ НАУКИCOMPUTER SCIENCEРАЗРАБОТКА RESIDUAL CNN-АРХИТЕКТУРЫ ДЛЯ РАСПОЗНАВАНИЯ ВЫРАЖЕНИЙ ЛИЦАDEVELOPMENT OF RESIDUAL CNN ARCHITECTURE FOR FACIAL EXPRESSION RECOGNITIONhttps://orcid.org/0009-0009-9353-7416АхметжанА.AkhmetkanA.

магистрант

г. Алматы

Master’s student

Almaty

az_akhmetkan@kbtu.kzhttps://orcid.org/0000-0002-1755-8161МуталиевЕ.MutaliyevYe.

докторант

г. Алматы

PhD student

Kaskelen

emutaliev11@gmail.comКазахстанско-Британский технический университетКазахстанKazakh-British Technical UniversityKazakhstanКазахстанско-Британский технический университетКазахстанSDU UniversityKazakhstan202629032026231163172Copyright © Ахметжан А., Муталиев Е., 20262026Ахметжан А., Муталиев Е.Akhmetkan A., Mutaliyev Y.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://vestnik.kbtu.edu.kz/jour/article/view/2510

Данное исследование представляет систему глубокой нейронной сети, которая обеспечивает многоклассовую классификацию эмоций в процессе разработки. Система определяет семь эмоциональных состояний: гнев (angry), отвращение (disgust), страх (fear), радость (happy), нейтральное состояние (neutral), грусть (sad) и удивление (surprise). Исследователи разделили свой набор данных на обучающую и тестовую части после предварительной обработки и использовали точность (precision), полноту (recall), F1-меру, матрицу ошибок (confusion matrix) и кривые ROC-AUC для оценки результатов. Согласно матрице ошибок, модель достигает наибольшей точности при распознавании радости (89%), далее следуют удивление (68%) и отвращение (49%). Модель показала хорошие и отличные результаты по большинству эмоций, однако испытывает трудности с эмоциями «страх» и «нейтральное состояние», поскольку их признаки пересекаются или распределение классов несбалансированно. Исследователи вычислили кривые операционных характеристик приемника (ROC) и площади под кривой (AUC) для каждого класса. Модель показала лучшие AUC-результаты для радости (0,92) и удивления (0,90), далее для отвращения (0,84). Наименьший AUC-показатель – 0,71 был зафиксирован для категории «страх» из-за слабой различимости этой эмоции. При оценке всех классов вместе модель достигла макроусредненного AUC-показателя 0,82. Предложенная нейронная сеть демонстрирует высокую эффективность в задачах распознавания эмоций благодаря своей способности выявлять интенсивные эмоции, такие как радость и удивление.

The research introduces a deep neural network system which achieves multi-class emotion classification through its development process. The system identifies seven emotional states through its classification system which includes angry, disgust, fear, happy, neutral, sad and surprise. The researchers divided their dataset into training and testing parts after preprocessing and they used precision and recall and F1-score and confusion matrix and ROC-AUC curves to evaluate their results. The model achieves its highest accuracy when detecting happy emotions at 89% followed by surprise at 68% and disgust at 49% according to the confusion matrix. The model achieves good to excellent classification results for most emotions yet it struggles with “fear and neutral emotions because their features overlap or their class distributions are unbalanced. The researchers computed Receiver Operating Characteristic (ROC) curves and Area Under Curve (AUC) values for each class in the study. The model produced its best AUC results for happy and surprise emotions at 0.92 and 0.90 respectively followed by disgust at 0.84. The lowest AUC score of 0.71 appeared in the fear category because this emotion showed weak discriminative properties. The model achieved a macro-averaged AUC score of 0.82 when evaluating all classes together. The proposed neural network shows strong performance in emotion recognition tasks through its ability to detect intense emotions such as happiness and surprise.

распознавание эмоцийглубокое обучениеостаточная сверточная нейронная сетьмногоклассовая классификацияанализ мимики лицаemotion recognitiondeep learningresidual convolutional neural networkmulti-class classificationfacial expression analysisReferencesEkundayo, O.S., and Viriri, S. Facial expression recognition: A review of trends and techniques. IEEE Access, 9, 136944–136973 (2021). https://doi.org/10.1109/ACCESS.2021.3113464Ekundayo, O.S., and Viriri, S. Facial expression recognition: A review of trends and techniques. IEEE Access, 9, 136944–136973 (2021). https://doi.org/10.1109/ACCESS.2021.3113464Sajjad, M., Ullah, F.U.M., Ullah, M., Christodoulou, G., Cheikh, F.A., Hijji, M., and Rodrigues, J.J. A comprehensive survey on deep facial expression recognition: challenges, applications, and future guidelines. Alexandria Engineering Journal, 68, 817–840 (2023). https://doi.org/10.1016/j.aej.2023.01.017Sajjad, M., Ullah, F.U.M., Ullah, M., Christodoulou, G., Cheikh, F.A., Hijji, M., and Rodrigues, J.J. A comprehensive survey on deep facial expression recognition: challenges, applications, and future guidelines. Alexandria Engineering Journal, 68, 817–840 (2023). https://doi.org/10.1016/j.aej.2023.01.017Zholshiyeva, L., Zhukabayeba, T., Serek, A.G., Duisenbek, R., Berdieva, M., and Shapay, N. Deep learning-based continuous sign language recognition. Journal of Robotics and Control (JRC), 6 (3), 1106–1118 (2025). https://doi.org/10.18196/jrc.v6i3.25881Zholshiyeva, L., Zhukabayeba, T., Serek, A.G., Duisenbek, R., Berdieva, M., and Shapay, N. Deep learning-based continuous sign language recognition. Journal of Robotics and Control (JRC), 6 (3), 1106–1118 (2025). https://doi.org/10.18196/jrc.v6i3.25881Zholshiyeva, L., Zhukabayeva, T., Baumuratova, D., and Serek, A.G. Design of QazSL sign language recognition system for physically impaired individuals. Journal of Robotics and Control (JRC), 6 (1), 191–201 (2025). https://doi.org/10.18196/jrc.v6i1.23879Zholshiyeva, L., Zhukabayeva, T., Baumuratova, D., and Serek, A.G. Design of QazSL sign language recognition system for physically impaired individuals. Journal of Robotics and Control (JRC), 6 (1), 191–201 (2025). https://doi.org/10.18196/jrc.v6i1.23879Serek, A., Amirgaliyev, B., Li, R.Y.M., Zhumadillayeva, A., and Yedilkhan, D. Crowd density estimation using enhanced multi-column convolutional neural network and adaptive collation. IEEE Access (2025). https://doi.org/10.1109/ACCESS.2025.3597393Serek, A., Amirgaliyev, B., Li, R.Y.M., Zhumadillayeva, A., and Yedilkhan, D. Crowd density estimation using enhanced multi-column convolutional neural network and adaptive collation. IEEE Access (2025). https://doi.org/10.1109/ACCESS.2025.3597393Singh, R., Saurav, S., Kumar, T., Saini, R., Vohra, A., and Singh, S. Facial expression recognition in videos using hybrid CNN & ConvLSTM. International Journal of Information Technology, 15 (4), 1819–1830 (2023). https://doi.org/10.1007/s41870-023-01183-0Singh, R., Saurav, S., Kumar, T., Saini, R., Vohra, A., and Singh, S. Facial expression recognition in videos using hybrid CNN & ConvLSTM. International Journal of Information Technology, 15 (4), 1819–1830 (2023). https://doi.org/10.1007/s41870-023-01183-0Pise, A.A., Alqahtani, M.A., Verma, P., K., P., Karras, D.A., S., P., and Halifa, A. Methods for facial expression recognition with applications in challenging situations. Computational Intelligence and Neuroscience, 2022 (1), 9261438 (2022). https://doi.org/10.1155/2022/9261438Pise, A.A., Alqahtani, M.A., Verma, P., K., P., Karras, D.A., S., P., and Halifa, A. Methods for facial expression recognition with applications in challenging situations. Computational Intelligence and Neuroscience, 2022 (1), 9261438 (2022). https://doi.org/10.1155/2022/9261438Ge, H., Zhu, Z., Dai, Y., Wang, B., and Wu, X. Facial expression recognition based on deep learning. Computer Methods and Programs in Biomedicine, 215, 106621 (2022). https://doi.org/10.1016/j.cmpb.2022.106621Ge, H., Zhu, Z., Dai, Y., Wang, B., and Wu, X. Facial expression recognition based on deep learning. Computer Methods and Programs in Biomedicine, 215, 106621 (2022). https://doi.org/10.1016/j.cmpb.2022.106621Dishar, H.K., and Muhammed, L.A. A review of the overfitting problem in convolution neural network and remedy approaches. Journal of Al-Qadisiyah for Computer Science and Mathematics, 15 (2), 155 (2023). https://doi.org/10.29304/jqcm.2023.15.2.1240Dishar, H.K., and Muhammed, L.A. A review of the overfitting problem in convolution neural network and remedy approaches. Journal of Al-Qadisiyah for Computer Science and Mathematics, 15 (2), 155 (2023). https://doi.org/10.29304/jqcm.2023.15.2.1240Bejani, M.M., and Ghatee, M. A systematic review on overfitting control in shallow and deep neural networks. Artificial Intelligence Review, 54 (8), 6391–6438 (2021).Bejani, M.M., and Ghatee, M. A systematic review on overfitting control in shallow and deep neural networks. Artificial Intelligence Review, 54 (8), 6391–6438 (2021).Salehi, A.W., Khan, S., Gupta, G., Alabduallah, B.I., Almjally, A., Alsolai, H., et al. A study of CNN and transfer learning in medical imaging: advantages, challenges, future scope. Sustainability, 15 (7), 5930 (2023). https://doi.org/10.3390/su15075930Salehi, A.W., Khan, S., Gupta, G., Alabduallah, B.I., Almjally, A., Alsolai, H., et al. A study of CNN and transfer learning in medical imaging: advantages, challenges, future scope. Sustainability, 15 (7), 5930 (2023). https://doi.org/10.3390/su15075930Gupta, J., Pathak, S., and Kumar, G. Deep learning (CNN) and transfer learning: a review. Journal of Physics: Conference Series, 2273 (1), 012029 (2022).Gupta, J., Pathak, S., and Kumar, G. Deep learning (CNN) and transfer learning: a review. Journal of Physics: Conference Series, 2273 (1), 012029 (2022).Nan, F., Jing, W., Tian, F., Zhang, J., Chao, K.M., Hong, Z., and Zheng, Q. Feature super-resolution based facial expression recognition for multi-scale low-resolution images. Knowledge-Based Systems, 236, 107678 (2022). https://doi.org/10.1016/j.knosys.2021.107678Nan, F., Jing, W., Tian, F., Zhang, J., Chao, K.M., Hong, Z., and Zheng, Q. Feature super-resolution based facial expression recognition for multi-scale low-resolution images. Knowledge-Based Systems, 236, 107678 (2022). https://doi.org/10.1016/j.knosys.2021.107678Lo, L., Ruan, B.K., Shuai, H.H., and Cheng, W.H. Modeling uncertainty for low-resolution facial expression recognition. IEEE Transactions on Affective Computing, 15 (1), 198–209 (2023). https://doi.org/10.1109/TAFFC.2023.3264719Lo, L., Ruan, B.K., Shuai, H.H., and Cheng, W.H. Modeling uncertainty for low-resolution facial expression recognition. IEEE Transactions on Affective Computing, 15 (1), 198–209 (2023). https://doi.org/10.1109/TAFFC.2023.3264719Pham, L., Vu, T.H., and Tran, T.A. Facial expression recognition using residual masking network. Proc. 25th Int. Conf. Pattern Recognition (ICPR), 4513–4519 (2021). https://doi.org/10.1109/ICPR48806.2021.9411919Pham, L., Vu, T.H., and Tran, T.A. Facial expression recognition using residual masking network. Proc. 25th Int. Conf. Pattern Recognition (ICPR), 4513–4519 (2021). https://doi.org/10.1109/ICPR48806.2021.9411919Roy, A.K., Kathania, H.K., Sharma, A., Dey, A., and Ansari, M.S.A. ResEmoteNet: bridging accuracy and loss reduction in facial emotion recognition. IEEE Signal Processing Letters (2024). https://doi.org/10.1109/LSP.2024.3521321Roy, A.K., Kathania, H.K., Sharma, A., Dey, A., and Ansari, M.S.A. ResEmoteNet: bridging accuracy and loss reduction in facial emotion recognition. IEEE Signal Processing Letters (2024). https://doi.org/10.1109/LSP.2024.3521321Bah, I., and Xue, Y. Facial expression recognition using adapted residual based deep neural network. Intelligence & Robotics, 2 (1), 72–88 (2022). https://doi.org/10.20517/ir.2021.16Bah, I., and Xue, Y. Facial expression recognition using adapted residual based deep neural network. Intelligence & Robotics, 2 (1), 72–88 (2022). https://doi.org/10.20517/ir.2021.16msambare. fer2013 (Facial Expression Recognition 2013 Dataset). Kaggle. URL: https://www.kaggle. com/datasets/msambare/fer2013 (accessed: 26.09. 2025).msambare. fer2013 (Facial Expression Recognition 2013 Dataset). Kaggle. URL: https://www.kaggle. com/datasets/msambare/fer2013 (accessed: 26.09. 2025).

The authors declare that there are no conflicts of interest present.