A Secure Web-based Application for Predicting Parkinson Disease using Speech Signal Features and Machine Learning Techniques

Parkinson’s disease (PD) is a progressive neurodegenerative disorder affecting motor control, including the muscle responsible for speech. Alterations in voice, such as reduced vocal stability and amplitude variation often appear in early and mid-stages of PD. Speech feature extraction provides a non-invasive, cost-effective, and scalable approach for remote screening of PD using minimal equipment. This study presents a secure, web-based application that predicts whether an individual has PD using speech recordings. Vocal biomarkers are extracted and used to train multiple binary classifiers, including Support Vector Machine, Logistic Regression, Random Forest, K-Nearest Neighbors, CatBoost, XGBoost, AdaBoost and Multiplayer Perception. Model outputs are combined using an ensemble based Stacking approach, achieving an overall accuracy of 95%. The application allows users to upload speech samples and instantly receive a PD likelihood prediction, making remote and accessible screening possible. This demonstrates the potential of speech-based machine learning systems for reliable Parkinson’s detection.