Parkinson's disease (PD) and essential tremor (ET) are movement disorders challenging to diagnose due to overlapping symptoms. This study investigates the use of automatic speech analysis to differentiate between PD and ET, addressing the challenge of cross-language variability. A Gaussian mixture model (GMM)-universal background model (UBM) approach, combined with support vector machines (SVM), was used to adapt models trained on German and Spanish speech data to classify Czech speakers. Results showed high accuracy (up to 86.2%) in differentiating PD and ET in binary classification, and moderate accuracy (up to 71.6%) in tri-class classification (including healthy controls). Prosody and articulation were identified as key differentiating speech features.
Publisher
npj Digital Medicine
Published On
Feb 17, 2024
Authors
Cristian David Rios-Urrego, Jan Rusz, Juan Rafael Orozco-Arroyave
Tags
Parkinson's disease
essential tremor
automatic speech analysis
Gaussian mixture model
support vector machines
speech features
cross-language variability
Related Publications
Explore these studies to deepen your understanding of the subject.
