Development of an AI-Driven System for Detecting Parkinson’s Disease Utilizing Patient Voice

The Parkinson disease (PD) is a persistent neurological disorder which impacts the motor system, resulting in variety of motor and other signs and symptoms. Individuals diagnosed with illness experience motor dysfunctions such as impairments in voice, writing, and ambulation. Voice analysis is a non-invasive way to diagnose PD. Consequently, researchers are increasingly using speech analysis models to identify and remotely monitor Parkinson’s disease.This work focus on pitch, jitter, shimmer, and formant frequencies for analysis of PD. We have utilized these parameters for machine learning and deep learning models Stack LR, Support Vector Machine (SVM), Decision Tree (DT), and Random Forest classifiers, CNN, LSTM, and GRU to evaluate Parkinson’s disease.We have used 195 audio recordings of 31 samples taken from UCI. The study reveals that CNN and stacking models with Decision Tree base models yield the best results with an overall accuracy of 95.11% to detect PD. The findings show that proposed approach performs better than most advanced techniques which can help doctors to diagnose PD at early stage.