• Botsa Akanksha
• Bharati Bidikar

Effective mental health interventions depend on the precise and timely assessment of depression severity. Earlier research predominantly relied on unimodal systems, focusing either on textual content or acoustic signals for depression detection. Although multimodal fusion approaches have been explored recently, the effectiveness of each modality individually remains significant for scalable applications. In this study, we propose a dual unimodal framework where separate models are developed for text and audio, each designed to assess depression severity independently. For the textual modality, a pre-trained BERT model is fine-tuned to classify Reddit posts into four categories: normal, mild, moderate, and severe. A rule-based labeling mechanism is employed to annotate unlabeled posts based on linguistic indicators of depressive symptoms. The BERT-based model achieved a high validation accuracy of 97.16%, demonstrating its ability to capture semantic patterns associated with different severity levels. In parallel, the audio-based pipeline utilized Mel Frequency Cepstral Coefficients (MFCCs) extracted from voice recordings representing three severity classes. A lightweight neural network was trained on these features under constrained conditions involving label noise and limited data availability. Despite these challenges, the model achieved an accuracy of 95% on noise-free test data, validating the reliability of acoustic features in depression detection. The outcomes from both approaches demonstrate that implementing automated, scalable, and non-invasive mental health screening instruments is feasible. The proposed separate unimodal models can serve as a foundation for future fusion-based multimodal systems, supporting clinical assessments and early diagnosis in real-world applications.

BERT, MFCC, Text Classification, Audio Classification, Deep Learning, Speech Processing, Transformer Models

"Automatic Detection of Depression Severity from Text and Audio using Deep learning", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 7, page no.f559-f564, July-2025, Available :http://www.jetir.org/papers/JETIR2507560.pdf

"Automatic Detection of Depression Severity from Text and Audio using Deep learning", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 7, page no. ppf559-f564, July-2025, Available at : http://www.jetir.org/papers/JETIR2507560.pdf