An Analysis of Vocal Features for Parkinson’s Disease Classification Using Machine Learning

The aging population continues to grow globally, which causes more people affected by Parkinson’s disease (PD). Unfortunately, underdeveloped regions face significant challenges in the timely and accurate detection of PD due to limited resources and awareness. Additionally, PD symptoms might be subtle at first and can vary greatly from patient to patient. This study proposes an innovative approach to address these issues by integrating multiple symptoms, including rest tremor and voice degradation, through smartphone-based data collection and cloud-enabled machine learning systems. The proposed system captures data using smartphone accelerometers and microphones, collecting information from both PD sufferers and healthy people. The data is then utilized to train and optimize high-performance machine learning models. Subsequently, the system is tested on new data from individuals suspected of having PD. By leveraging majority voting across trained algorithms, the system identifies PD cases and connects detected patients with nearby neurologists for consultation. This approach aims to enhance early PD detection and management, particularly in resource-limited settings, by utilizing accessible and scalable technologies.