SUDA LAGUNADA SAI VENKATA, SHAIK ABRRAR, SAMMITI BHARATH KUMAR, IRLA VENKATA DURGA PRASAD, PARSUBOINA NAVEEN, GYOTHI CHOWDARY

DOI: https://doi.org/10.5281/zenodo.23257888

Psychological trauma often remains unnoticed because it is subtle and rarely monitored continuously, and manual assessment is time-consuming and subjective. This paper describes the design of an AI-based web application that analyzes user-supplied voice and text for possible trauma-related indicators. Text is analyzed with Natural Language Processing (NLP) techniques, while voice is processed with speech analysis to extract emotional characteristics such as tone, pitch and speech patterns. The two sources are combined in a multimodal approach, and the user’s state is classified as Low, Medium or High risk. When high risk is identified, the system lets the user contact a healthcare professional by call or messaging. The system also includes registration and login, report generation, database storage and an administrative module. The aim is early awareness, accessibility and timely support. The project report documents the design, requirements and implementation approach; it does not report measured accuracy or clinical validation. The system is a decision-support tool and is not intended to replace professional diagnosis

trauma detection, mental health, natural language processing, speech emotion recognition, multimodal analysis, risk classification, healthcare recommendation, web application

This paper described an AI-based web application that analyzes voice and text with NLP and speech processing, fuses them for Low/Medium/High risk classification, and connects high-risk users with healthcare professionals. The contribution is an integrated design, with requirements and architecture, for early awareness and accessible support. Quantitative evaluation and clinical validation remain future work, and the system is not a replacement for professional diagnosis.

R. A. Calvo and D. N. Milne, “Natural language processing in mental healthapplications using non-clinical texts,” Nat. Lang. Eng., vol. 22, no. 5, pp. 649–685, 2016.
[2] A. S. Miner, A. Milstein, J. T. Hancock et al., “Talking to machines about personal mental health problems,” JAMA, vol. 318, no. 13, pp. 1217–1218, 2017.
[3] N. Cummins, S. Scherer, J. Krajewski et al., “A review of depression and suicide risk assessment using speech analysis,” Speech Commun., vol. 71, pp. 10–49, 2015.
[4] L. A. Low, N. C. Maddage, M. Lech et al., “Influence of acoustic low-level descriptors in the detection of clinical depression in adolescents,” IEEE Trans. Biomed. Eng., vol. 58, no. 3, pp. 574–586, 2011. 3
[5] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training ofdeep bidirectional transformers for language understanding,” in Proc. NAACL-HLT, 2019.
[6] J. W. Pennebaker, M. R. Mehl, and K. G. Niederhoffer, “Psychologicalaspects of natural language use,” Annu. Rev. Psychol., vol. 54, pp. 547–577, 2003.
[7] S. Scherer, G. Stratou, J. Gratch, and L.-P. Morency, “Investigating voicequality as a speaker-independent indicator of depression and PTSD,” in Proc. Interspeech, 2013.
[8] M. E. Larsen and T. W. Boonstra, “Detecting depression from social mediatext using deep learning,” IEEE J. Biomed. Health Inform., vol. 22, no. 3, pp. 1 10, 2018.
[9] B. Schuller and S. Steidl, “Speech emotion recognition using deep learningtechniques,” IEEE Trans. Affective Comput., vol. 10, no. 4, pp. 1–12, 2019.
[10] S. Poria and E. Cambria, “Multimodal emotion recognition using deeplearning,” ACM Trans. Multimedia Comput., vol. 15, no. 1, pp. 1–19, 2019.
[11] K. Fitzpatrick and A. Darcy, “Conversational agents for mental healthapplications,” J. Med. Internet Res., vol. 19, no. 2, pp. 1–10, 2017.
[12] A. Kumar and D. Singh, “AI-based web application for real-time mentalhealth monitoring using speech and text analytics,” Int. J. Med. Inform., vol. 169, Art. 104933, 2023.
[13] T. Al Hanai, M. Ghassemi, and J. Glass, “Multimodal depression detectionusing deep learning,” in Proc. Interspeech, 2018, pp. 1–5.
[14] S. Ji and X. Pan, “MentalBERT: Pretrained language models for mentalhealthcare,” ACL Findings, 2021.
[15] J. Gratch et al., “The Distress Analysis Interview Corpus (DAIC-WOZ): A dataset for depression detection,” World Psychiatry, vol. 10, no. 3, 2019.
[16] J. Torous and M. Wisniewski, “Digital mental health interventions: Opportunities and challenges,” World Psychiatry, vol. 19, no. 3, pp. 1–12, 2020.
[17] M. Neumann and N. T. Vu, “Speech emotion recognition for traumadetection using deep learning,” IEEE Trans. Affective Comput., vol. 9, no. 2, pp. 1–12, 2018.
[18] G. Coppersmith and M. Dredze, “Natural language processing for mentalhealth detection,” in Proc. NAACL-HLT, 2015, pp. 1–10.
[19] Z. Zeng and M. Pantic, “Multimodal emotion recognition using speech andtext data,” IEEE Signal Process. Mag., vol. 26, no. 3, pp. 1–15, 2019.
[20] S. Pan and J. Huang, “Deep learning-based stress detection using behavioraldata,” IEEE Access, vol. 8, pp. 1–12, 2020.
[21] R. Sharma and P. Gupta, “AI-based doctor recommendation system formental health support,” Int. J. Healthcare Inform., vol. 12, no. 2, pp. 45–60, 2022.
[22] A. Kumar and N. Verma, “AI-based doctor recommendation system fortrauma care,” Int. J. Smart Healthcare, vol. 3, no. 2, pp. 41–50, 2024.

SUDA LAGUNADA SAI VENKATA, SHAIK ABRRAR, SAMMITI BHARATH KUMAR, IRLA VENKATA DURGA PRASAD, PARSUBOINA NAVEEN, GYOTHI CHOWDARY (2026). Detecting Trauma Symptoms and Connecting Doctors Using Voice and Text Analysis. International Journal of Computer Science Engineering Techniques, 10(5), 257–261. ISSN: 2455-135X. DOI: https://doi.org/10.5281/zenodo.23257888