Parkinson’s Disease Detection by Processing Different ANN Architecture Using Vocal Dataset

Abstract

Parkinson's Disease (PD) is a long-standing neurodegenerative condition of the central nervous system that mainly affects the motor system and origins full or partial damage in behavior, speech, motor reflexes, mental processing, and other energetic functions. Doctors use different types of datasets such as speech, movement and images from the people to diagnose the disease. In this paper, the speech dataset is collected from people with and without PD to detect the disease. The voice recording samples are analyzed and the feature vectors are extracted from the voice samples. A supervised ANN Multi-Layer Perceptron with a backpropagation algorithm is presented to accurately diagnose and distinguish between healthy and PD individuals. Different Architecture with diverse neuron numbers in the hidden layers are tested to utilize the model and the result of each architecture is compared to select the best ANN architecture for PD recognition. So far, our model score is the highest which is 93% for the testing dataset.