Research papers
Peer-reviewed research and preprints in machine learning and AI
Journal of Voice — peer-reviewed article, accepted 8 October 2024 • 2024
Mental Health Diagnosis From Voice Data Using Convolutional Neural Networks and Vision Transformers
A hybrid CNN–Vision Transformer approach for classifying Bengali voice samples into stable and unstable mental-health categories using spectrogram representations. The study reports approximately 91% accuracy and ROC-AUC of about 0.97, with a focus on ethical data collection and research evaluation rather than clinical validation.
arXiv preprint arXiv:2501.18161 • 2025
Using Computer Vision for Skin Disease Diagnosis in Bangladesh: Enhancing Interpretability and Transparency in Deep Learning Models for Skin Cancer Classification
A skin-lesion classification study using HAM10000, a custom deep convolutional network, and established transfer-learning baselines. The work emphasizes interpretable decision support and the Bangladesh healthcare context. This item is a preprint and is presented separately from the peer-reviewed journal article.
arXiv preprint arXiv:2601.16793 — submitted 23 January 2026 • 2026
A Novel Transfer Learning Approach for Mental Stability Classification from Voice Signal
A transfer-learning study for mental-stability classification from voice spectrograms under limited-data conditions. VGG16, InceptionV3, and DenseNet121 were evaluated across non-augmented, augmented, and transfer-learning phases with strict data separation. DenseNet121 achieved the strongest reported result: 94% accuracy and 99% AUC.