Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM62858
Volume:
10 |
5 |
May 2026
Pages:
1-7
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Department of Computer Science and Engineering
JAIN (Deemed-to-be-University)
Department of Computer Science and Engineering
JAIN (Deemed-to-be-University)
Department of Computer Science and Engineering
JAIN (Deemed-to-be-University)
Department of Computer Science and Engineering
JAIN (Deemed-to-be-University)
Department of Computer Science and Engineering
JAIN (Deemed-to-be-University)
The global mental health treatment gap exceeds 75% in developing nations, driven by cost, stigma, and geographic inaccessibility to qualified practitioners. Existing digital mental health tools are predominantly text-input-centric, demanding that distressed users articulate complex emotions through a keyboard at precisely the moments when cognitive load is highest. This paper presents VoiceMind, a voice-first agentic AI wellness companion implemented as a cross-platform Flutter mobile application backed by a FastAPI server on Google Cloud. The contribution is a reference architecture integrating five components into a single mobile stack: (1) a dual-channel emotion fusion design combining 71 acoustic biomarkers extracted via librosa with Gemini 2.5 Flash text inference at a fixed 60/40 acoustic-dominant weighting; (2) a tiered crisis detection architecture operating at client (Dart regex), REST (Python regex), and WebSocket (turn accumulator) layers, deterministically matching 34 explicit crisis-indicator phrases; (3) a multi-agent therapeutic pipeline using Google’s Agent Development Kit (ADK), routing conversations between Triage, Therapist (CBT/DBT-informed), and Crisis de-escalation agents; (4) a keyword-matched offline coping engine providing local fallback across 10 emotional categories, 20 coping tools, and 20 wellness activities without network access; and (5) a voice-first turn-based conversational loop coupling device speech-to-text and text-to-speech with server-side acoustic analysis on every turn. The implementation passes 160/160 Flutter unit tests and 85/85 backend pytest tests with zero lint issues. The Random Forest classifier is currently trained on synthetic acoustic feature profiles; training on real speech corpora, latency benchmarking under load, and a human-subjects user study are identified as primary future work.
KEYWORDS
The complete codebase is validated through an automated test suite, 160/160 Flutter unit tests and 85/85 backend pytest tests pass with zero lint issues, establishing that the system behaves as specified across all integration points described in Section III. The contribution of this work is architectural and implementation-level, not clinical: the Random Forest classifier is currently trained on synthetic acoustic feature profiles; the system has not been evaluated on real emotional speech, on production load, or with human participants.
[1] World Health Organization, “Mental health: strengthening our response,” WHO Fact Sheet, 2023. [2] M. Marie et al., “Acoustic biomarkers of psychological distress: A 2025 PRISMA systematic review,” IEEE Trans. Affective Computing, 2025. [3] S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan, and Y. Cao, “ReAct: Synergizing reasoning and acting in language models,” in Proc. Int. Conf. Learning Representations (ICLR), 2023. [4] G. N. Peerzade, R. Deshmukh, and S. D. Waghmare, “A review on speech emotion recognition,” Int. J. Computer Sciences and Engineering, vol. 6, no. 3, pp. 400–402, 2018. [5] D. D. Joshi and M. B. Zalte, “Speech emotion recognition: A review,” IOSR J. Electronics and Communication Engineering, vol. 4, no. 4, pp. 34–37, 2013. [6] K. Chlasta, K. Wołk, and I. Krejtz, “Automated speech-based screening of depression using deep convolutional neural networks,” Procedia Computer Science, vol. 164, pp. 618–628, 2019. [7] E. Lieskovská, M. Jakubec, R. Jarina, and M. Chmulík, “A review on speech emotion recognition using deep learning and attention mechanism,” Electronics, vol. 10, no. 10, p. 1163, 2021. [8] L. Pepino, P. Riera, and L. Ferrer, “Emotion recognition from speech using wav2vec 2.0 embeddings,” in Proc. Interspeech, 2021, pp. 3400–3404. [9] R. A. Khalil, E. Jones, M. I. Babar, T. Jan, M. H. Zafar, and T. Alhussain, “Speech emotion recognition using deep learning techniques: A review,” IEEE Access, vol. 7, pp. 117327–117345, 2019. [10] T. Baltrušaitis, C. Ahuja, and L.-P. Morency, “Multimodal machine learning: A survey and taxonomy,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 41, no. 2, pp. 423–443, 2018.
Publication Date
5/8/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
7
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...