Detection of Parkinson's Disease Using Vocal Features: An Eigen Approach

Parkinson's disease (PD) is a degenerative disorder of the human central nervous system that causes tremors and affects movement. Symptoms usually appear gradually over time. Researchers are seeking biomarkers for Parkinson's disease in the hopes of allowing for earlier detection and more tailored treatments to slow the disease's progression. Existing methods of diagnosis include Blood tests, MRI scans, and PET scans. However, these are highly time and resource-consuming. PD also shows an amble change in the voice patterns of a person. Hence, acoustic analysis of voice signals can indicate the progression of PD. This can be analysed using a trained classifier model, which provides an easy diagnosis of the disease. This paper analyses the performance of AI-ML models viz- Linear regression, Support Vector Machine (SVM), K-Nearest Neighbourhood (KNN), Ran-dom Forest, and XG Boost for the detection of Parkinson's disease using vocal feature sets. Experimental evaluations show that the Random Forest model produced an impressive accuracy of 100%. The classification algorithms' accuracy, precision, recall, F1-score, and Mathews Correlation Coefficient (MCC) are all examined. The Random Forest classifiers are 100% accurate, with an accuracy of 1.000, recall of 1.000, F1-score of 1.000, and MCC of 1.000. Implementing dimensionality reduction using the Eigen approach (Principal Component Analysis) and the dataset combination are the critical reasons for the reported high accuracy. The potential of this methodology is prominent as it can be used to diagnose various other diseases, such as asthma, cancer, and Alzheimer's disease.