Diagnosing Parkinson's Disease with KNN Classifier Utilizing Speech Feature Extraction
Srikrithi Santhanam, A. Advika, Srinithi Santhanam, J. Faritha Banu · 2024
A significant portion of people have suffered from a form of Parkinson's disease (PD), widely attributed to be the second most frequently diagnosed form of neurological illness that significantly impairs motor and cognitive abilities. But as common a disease as it is, a lack of proper conclusive testing results in diagnoses that rely on historical analysis and records and can lead to errors, particularly in the early stages. Using a dataset from the University of Oxford that includes a variety of speech variables recognised for their non-invasive and distinctive qualities in the identification of this neurological ailment, this study investigates a machine-learning probability of diagnostic tests. This method leverages K-Nearest Neighbours (KNN) classification, using the features: fundamental frequency parameters, jitter, shimmer, and other vocal disturbances. Voice analysis shows great promise for improving early identification and ensuring proper protection of Parkinson's victims from this haunting disease.