Early Detection of Parkinson Disease using Voice Data

Abstract

Parkinson’s disease affects over 10 million people worldwide, with approximately 20 percent of patients not being diagnosed. Clinical diagnosis is expensive because there are no specific tests or bio-markers, and it can take days to diagnose because it is based on a comprehensive evaluation of the individual’s symptoms. Existing research either predicts a Unified Parkinson Disease Rating Scale rating, uses other key Parkinsonian features to diagnose an individual, such as tapping, gait, and tremor, or focuses on different audio features. In this paper, we are focusing on using the voice aspect for the early detection of the disease. We use the University of California Irvine (UCI) Parkinson data set. This data set contains various parameters regarding voice jitter. The data set first undergoes preprocessing. We have used a Feedforward Neural Network (FNN) model to acquire early on detection using the above data set. Our model has achieved an efficiency of 97.43 percent. This efficiency can be improved by using even a larger and diverse data set.