Parkinson's disease (PD) is a progressive neurodegenerative disorder in which speech impair-ment appears at an early stage, making voice analysis a reliable and non-invasive biomarker for diagnosis and monitoring. However, many existing speech-based systems rely on manually engi-neered acoustic features and focus primarily on disease detection without providing objective severity assessment or continuous monitoring. This project presents an end-to-end AI-based real-time Parkinson's disease detection and monitoring system using voice analysis. Speech signals are transformed into log-Mel spectrogram representations to preserve spectral and temporal char-acteristics and are analyzed using a Conformer-based deep learning architecture that integrates convolutional layers with self-attention mechanisms. A multi-task learning strategy is employed to simultaneously identify Parkinson's disease and estimate its severity level. To improve robust-ness, data augmentation techniques are applied to handle noise and speaker variability. The sys-tem further incorporates an AI-driven recommendation module that provides personalized health suggestions based on the predicted results. A secure cloud- connected dashboard enables real-time storage, visualization, and tracking of patient data over time. Additional modules, including speech emotion analysis, smart alerts via email and SMS, automated PDF report generation, mobile application integration, and data encryption, enhance usability, safety, and accessibility. The proposed system supports early detection, objective disease monitoring, and scalable de-ployment in clinical and remote healthcare environments.
S et al. (Thu,) studied this question.