6

Insonning yuz tasviridan hissiyotlarni aniqlash inson-kompyuterning 
o‘zaro ta’siri, xavfsizlik tizimlari, ruhiy salomatlik monitoringi va aqlli tizimlar 
uchun muhim yo‘nalishlardan biri hisoblanadi. Ayniqsa, inson-kompyuterning 
o‘zaro ta’siri yo‘nalishida humanoid robotlar ishlab chiqishda robotning 
hissiyotlarni anglab, unga mos tarzda muloqot qilishi muhim ahamiyat kasb etadi. 
Shu bois hissiyotlarni aniqlash tizimlarini yaratish kompyuter ko‘rish va chuqur 
o‘rganishning muhim masalalaridan biridir. Ushbu tadqiqotda inson yuz tasviridan 
hissiyotlarni aniqlash uchun avvaldan o‘qitilgan mashhur konvolyutsion neyron 
tarmoqlarini yaxshilangan FER-2013 ma’lumotlar to‘plamida o‘qitish orqali 
samarali yondashuv taklif etildi. Tadqiqot davomida ilg‘or konvolyutsion neyron 
tarmoqlari yordamida hissiy holatni aniqlash vazifasi samaradorligi o‘rganilib, 
oldindan o‘qitilgan mashhur arxitekturalar – ResNet-50, VGGNet-16, DenseNet-121 
va EfficientNet-B0 modellarining yuz ifodalarini aniqlash bo‘yicha natijalari 
tahlil qilindi. Tadqiqotda FER-2013 ma’lumotlar to‘plamining yaxshilangan va 
to‘ldirilgan versiyasidan foydalanildi. Ma’lumotlarni qayta ishlash jarayonida yuz 
ifodalari orasidagi muvozanat buzilishi, sifati past rasmlar va noto‘g‘ri belgilangan 
(labeling) tasvirlar aniqlanib, ularni tuzatish ishlari amalga oshirildi. Shuningdek, 
overfitting muammosini kamaytirish maqsadida ma’lumotlarni kengaytirish 
(data augmentation) texnikalari qo‘llandi. Model natijalari aniqlik (accuracy) va 
yo‘qotish funksiyasi (loss) kabi mezonlar asosida baholandi.

  • Web Address
  • DOI
  • Date of creation in the UzSCI system 02-06-2025
  • Read count 6
  • Date of publication 14-04-2025
  • Main LanguageO'zbek
  • Pages43-52
Ўзбек

Insonning yuz tasviridan hissiyotlarni aniqlash inson-kompyuterning 
o‘zaro ta’siri, xavfsizlik tizimlari, ruhiy salomatlik monitoringi va aqlli tizimlar 
uchun muhim yo‘nalishlardan biri hisoblanadi. Ayniqsa, inson-kompyuterning 
o‘zaro ta’siri yo‘nalishida humanoid robotlar ishlab chiqishda robotning 
hissiyotlarni anglab, unga mos tarzda muloqot qilishi muhim ahamiyat kasb etadi. 
Shu bois hissiyotlarni aniqlash tizimlarini yaratish kompyuter ko‘rish va chuqur 
o‘rganishning muhim masalalaridan biridir. Ushbu tadqiqotda inson yuz tasviridan 
hissiyotlarni aniqlash uchun avvaldan o‘qitilgan mashhur konvolyutsion neyron 
tarmoqlarini yaxshilangan FER-2013 ma’lumotlar to‘plamida o‘qitish orqali 
samarali yondashuv taklif etildi. Tadqiqot davomida ilg‘or konvolyutsion neyron 
tarmoqlari yordamida hissiy holatni aniqlash vazifasi samaradorligi o‘rganilib, 
oldindan o‘qitilgan mashhur arxitekturalar – ResNet-50, VGGNet-16, DenseNet-121 
va EfficientNet-B0 modellarining yuz ifodalarini aniqlash bo‘yicha natijalari 
tahlil qilindi. Tadqiqotda FER-2013 ma’lumotlar to‘plamining yaxshilangan va 
to‘ldirilgan versiyasidan foydalanildi. Ma’lumotlarni qayta ishlash jarayonida yuz 
ifodalari orasidagi muvozanat buzilishi, sifati past rasmlar va noto‘g‘ri belgilangan 
(labeling) tasvirlar aniqlanib, ularni tuzatish ishlari amalga oshirildi. Shuningdek, 
overfitting muammosini kamaytirish maqsadida ma’lumotlarni kengaytirish 
(data augmentation) texnikalari qo‘llandi. Model natijalari aniqlik (accuracy) va 
yo‘qotish funksiyasi (loss) kabi mezonlar asosida baholandi.

Русский

 Определение эмоций по изображению лица человека является 
важным направлением во взаимодействии человек – компьютер, систем 
безопасности, мониторинга психического здоровья и интеллектуальных 
систем. Особенно в контексте взаимодействия человек – компьютер 
важным является разработка гуманоидных роботов, которые могут 
распознавать эмоции и взаимодействовать с человеком соответствующим 
образом. Поэтому создание систем определения эмоций является одной 
из ключевых задач компьютерного зрения и глубокого обучения. В данном 
исследовании предложен эффективный подход к определению эмоций 
на изображениях лица человека с помощью предварительно обученных 
популярных свёрточных нейронных сетей и улучшенной базы данных FER-
2013. В ходе исследования была оценена эффективность задачи определения 
эмоционального состояния с использованием передовых свёрточных 
нейронных сетей, а также проанализированы результаты по определению 
выражений лица для заранее обученных популярных архитектур: ResNet-50, 
VGGNet-16, DenseNet-121 и EfficientNet-B0. Для исследования использовалась 
улучшенная и дополненная версия базы данных FER-2013. В процессе обработки данных были выявлены и исправлены несоответствия в 
выражениях лиц, низкое качество изображений и ошибки в разметке (label-
ing). Также для снижения проблемы переобучения были применены техники 
увеличения данных (data augmentation). Результаты моделей были оценены 
по таким метрикам, как точность (accuracy) и функция потерь (loss).

English

 Emotion recognition from human facial images is one of the important 
directions for human-computer interaction, security systems, mental health 
monitoring, and intelligent systems. Especially in the development of humanoid 
robots in the field of human-computer interaction, one of the important features 
of a humanoid robot is that the robot can communicate with humans while sensing 
their emotions. The development of emotion recognition systems is a significant 
challenge in computer vision and deep learning. In this study, we suggest a good 
way to recognize emotions from people's facial images by using well-known 
convolutional neural networks that have been trained on a better version of 
the FER-2013 dataset. In this study, the effectiveness of implementing the task 
of emotional state recognition using advanced convolutional neural networks 
is studied. In particular, the results of the popular pre-trained architectures 
ResNet-50, VGGNet-16, DenseNet-121, and EfficientNet-B0 on facial expression 
recognition are analyzed. The study used an improved and augmented version 
of the FER-2013 dataset. During data processing, imbalances between facial 
expressions, low-quality images, and incorrectly labeled images were detected 
and corrected. In addition, data augmentation techniques were used to reduce the 
problem of overfitting. The model results were evaluated based on criteria such as 
accuracy and loss function.

№ Author name position Name of organisation
1 Kurbanov A. . tayanch doktorant Mirzo Ulug‘bek nomidagi O‘zbekiston Milliy Universiteti Jizzax filiali
№ Name of reference
1 Agung, E. R. (2024). Image-based facial emotion recognition using convolutional neural network on Emognition dataset. Sci Rep, 14.
2 Chouhayebi, H., & M. M. (2024). Human emotion recognition based on spatio-temporal facial features using HOG-HOF and VGG-LSTM. Computers, 101
3 Dada, E. G. (2023). Facial emotion recognition and classification using the Convolutional Neural Network-10 (CNN-10). Applied Computational Intelligence and Soft Computing, 2457898.
4 Ekman, P. (1978). Facial action coding systems. Consulting Psychologists Press.
5 Gursesli, M. (2024). Facial emotion recognition (FER) through custom lightweight CNN model: Performance evaluation in public datasets. IEEE Access, 1-1.
6 He, K., & X. Z. (2016). Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 770–778). Las Vegas, NV, USA. https://doi. org/10.1109/CVPR.2016.90
7 Simonyan, K. (2015). Very deep convolutional networks for large-scale image recognition. CoRR, abs/1409.1556. arXiv preprint arXiv:1409.1556.
8 Khaireddin, Y. (2021). Facial emotion recognition: State of the art performance on FER2013. https://doi.org/10.48550/arXiv.2105.03588
9 Kurbanov, A. A. (2024a). Inson tanasi harakatlarini tahlil qilishda zamonaviy modellar va algoritmlarni qo‘llashni o‘rganish [Studying the application of modern models and algorithms in analyzing human body movements]. (In Uzbek). The Descendants of Al-Farghani, 169–175.
10 Kurbanov, A. A. (2024b). Inson yuz tasviridan hissiyotlarni aniqlash uchun geometriya xususiyatlarini va tashqi ko‘rinishga asoslangan xususiyatlarini ajratib olish [Extraction of geometric features and appearance-based features for emotion detection from human facial images]. (In Uzbek). The Descendants of Al-Farghani, 61–67.
11 Mao, A. M. (2023). Cross-entropy loss functions: Theoretical analysis and applications. In International Conference on Machine Learning (pp. 23803-23828).
12 Minaee, S. (2021). Deep-emotion: Facial expression recognition using attentional convolutional network. Sensors, 30–46.
13 Oguine, O. C. (2022). Hybrid facial expression recognition (2013) model for real-time emotion classification and prediction. arXiv preprint arXiv:2206.09509.
14 Picard, R. W. (2000). Affective computing. Massachusetts Institute of Technology.
15 Tan, M. (2020). Ef�icientNet: Rethinking model scaling for convolutional neural networks. In International Conference on Machine Learning (pp. 6105–6114). PMLR.
Waiting