Early Parkinson's Disease Identification from Voice Recordings using Machine learning

Abstract

In many experimental studies by applying supervised and non-supervised learning methods it is being observed that the human voice progressively worsens with time. Parkinson’s Disease (PD) mostly affects the elderly more than 60 years old or more. PD is neurological disease that affects movements of the body and also causes tremors long time walking difficulty. And it is found by many researchers that voice disintegration is one of the earliest symptoms of PD diagnosis in which machine learning can be used to improve the accuracy of the diagnosis so that it can be detected early stage. Parkinson’s Disease have been very attractive to researchers so there is a lot of voice dataset available on UCI Machine Learning repository. We used Parkinson’s voice dataset which includes short sentences, numbers, words, and vowels which is compiled from PD patients and healthy subjects as well. Few researchers assert that concurrently capturing data from all participants using metrics for central tendency and dispersion is an effective method for creating PD prediction models. However, they failed to take into account that, in comparison to healthy people, PD patients have trouble pronouncing a small number of alphabets. Therefore, we will categorise each alphabet independently under this framework.. The final result would be majority vote from all the classifiers.