Detection of Parkinson's Disease using Machine Learning Algorithms from Speech Signals

Parkinson's disease is a long-term neurological con-dition that impacts 60 % of people above 50 years of age, it results in a decrease in voice in 90% of individuals and ranks as the second most prevalent illness among the elderly. It affects the gait, speech, and memory. The diagnosis and treatment requires physical visits which are extremely difficult and cumbersome for the older people. Making use of speech signals for the Parkinson's disease diagnosis provides a non-invasive, less time-consuming and painless experience to the patients. This paper's main objective is to build a feature vector that comprises of a time domain feature, Zero Crossing Rate, a frequency domain feature, the Spectral Centroid and the cepstral features from the 14 coefficients of Mel Frequency Cepstral Coefficients (MFCC), MFCC Delta and MFCC Delta-Delta. Support Vector Machine (Linear and Polynomial kernels), Random Forest, and Logistic Regression are utilised to classify Parkinson's disease. Support Vector Machine (both Linear and Polynomial kernels) achieved the maximum accuracy of 92.31 % as in contrast to the other machine learning algorithms considered in this paper. It is found from the comparison with the existing models that the proposed model with SVM classifier performed better than the other models.