Tremorsense - Parkinson's Detection

Abstract

I Parkinson's Disease (PD) Is A Progressive Neurodegenerative Disorder That Primarily Affects Motor Function. Early Diagnosis Is Crucial For Effective Management, Yet Current Methods Can Be Invasive, Costly, And Time-consuming. This Project, Tremorsense, Presents A Non-invasive, Accessible, And Efficient Solution For The Preliminary Detection Of Parkinson's Disease Using Vocal Biomarkers. The System Is Developed As A Web Application Where Users Can Record And Upload A Short Voice Sample. This Sample Is Then Processed To Extract A Range Of Acoustic Features Known To Be Affected By PD, Such As Jitter, Shimmer, And Fundamental Frequency Variations. A Pre-trained Random Forest Machine Learning Model Analyzes These Features To Classify The Sample And Provide A Risk Assessment Score. The Methodology Follows A Standard Data Science Workflow, Including Data Preprocessing, Feature Extraction, And Model Training On A Publicly Available Dataset Of Voice Samples From Healthy Individuals And PD Patients. The Resulting Web Application Provides An Intuitive User Interface, Ensuring Ease Of Use For Individuals Without Technical Expertise. The Core Technologies Used Are Python For The Backend, The Scikit-learn Library For The Machine Learning Model, And HTML/CSS/JavaScript For The Frontend. Black Box Testing Was Conducted To Ensure Functionality And Usability. The Research Results In A Functional Prototype That Can Help Individuals Receive An Early Indication Of Risk, Encouraging Them To Seek Professional Medical Advice Sooner And Demonstrating The Potential Of Machine Learning In Modern Diagnostics.

Keywords