Authors : V. Kiruthiga; K. Lakshmi Priya

Volume/Issue : Volume 10 - 2025, Issue 10 - October

Google Scholar : https://tinyurl.com/2u7hmmes

Scribd : https://tinyurl.com/bdhmtymt

DOI : https://doi.org/10.38124/ijisrt/25oct657

Note : A published paper may take 4-5 working days from the publication date to appear in PlumX Metrics, Semantic Scholar, and ResearchGate.

Note : Google Scholar may take 30 to 40 days to display the article.

Abstract : Mental health problems like depression and anxiety are increasing all over the world. Detecting them early can help people get proper care and support. Artificial Intelligence (AI) systems can analyze how people speak, write, or express emotions to find early signs of these problems. This study compares two types of learning methods — unimodal (using one type of data such as text or voice) and multimodal (using more than one type, like text, voice, and facial expressions). Both methods are tested using privacy-aware AI techniques such as Federated Learning and Differential Privacy, which protect user data from being shared or misused. The system was tested on public datasets like DAIC-WOZ and WESAD. The results show that multimodal learning gives better accuracy (about 10–12% higher) than unimodal learning, but it also needs more processing power and care to protect privacy. This comparison helps researchers understand the balance between accuracy, privacy, and efficiency when designing AI tools for mental health support.

Keywords : Multimodal Learning, Unimodal Learning, Mental Health Prediction, Privacy-Aware AI, Federated Learning, Differential Privacy, Ethical AI.

References :

1. F. Ringeval et al., “AVEC 2019 Workshop and Challenge: State-of-Mind, Depression, and Cross-Cultural Affect Recognition,” Proc. AVEC, 2019.
2. M. Valstar et al., “Detection of Depression from Facial Expressions, Audio and Text: AVEC 2016 Challenge,” Proc. AVEC, 2016.
3. H. Lin, J. Qiu, and S. Li, “Text-Based Depression Detection Using Transformer Language Models,” IEEE Transactions on Affective Computing, vol. 12, no. 4, pp. 957–968, 2021.
4. J. Han, Z. Zhang, and B. Schuller, “Privacy-Preserving Speech Emotion Recognition Using Secure Feature Representations,” Proc. IEEE ICASSP, pp. 6319–6323, 2022.
5. S. Poria, E. Cambria, D. Hazarika, and N. Majumder, “Multimodal Sentiment Analysis: Addressing Key Issues and Challenges,” IEEE Intelligent Systems, vol. 35, no. 6, pp. 17–25, 2020.
6. A. Kaissis, M. R. Makowski, D. Rückert, and R. F. Braren, “Secure, Privacy-Preserving and Federated Machine Learning in Medical Imaging,” Nature Machine Intelligence, vol. 2, no. 6, pp. 305–311, 2020.
7. Y. Liu, H. Wu, and J. Zhang, “Federated Learning for Mental Health Prediction: Opportunities and Challenges,” Proc. IEEE BHI, 2021.
8. Z. Zhao, G. Li, and L. Zhang, “A Review of Multimodal Depression Detection: Methods and Datasets,” Frontiers in Psychology, vol. 13, no. 921456, pp. 1–12, 2022.
9. T. Baltrušaitis, C. Ahuja, and L.-P. Morency, “Multimodal Machine Learning: A Survey and Taxonomy,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 41, no. 2, pp. 423–443, 2019.
10. P. Kairouz et al., “Advances and Open Problems in Federated Learning,” Foundations and Trends in Machine Learning, vol. 14, no. 1–2, pp. 1–210, 2021.