Effectiveness of Speech Analysis in Classification of Neurodegenerative Diseases: A Study on Parkinson's Disease

Neurodegenerative (ND) diseases have caused significant increase in the death rates across the world. The urgent need for biomarkers for diseases like Alzheimer's disease (AD), Parkinson's disease (PD) and amyotrophic lateral sclerosis (ALS), is vastly growing area of research in biomedical engineering. Speech production is a complex neuromuscular task that get affected due to ND diseases. In this study, we investigated the effectiveness of speech analysis in the classification of ND diseases focusing on PD. In the literature, existing methodologies in speech analysis using signal processing aim at extraction and usage of features to capture perturbations due to PD. A novel methodology is proposed in this paper that extracts features at a pitch synchronous level from the speech. Instead of the average, variance in these features at a syllabic level is used for classification. An unsupervised k-means clustering algorithm is used to perform classification. Feature selection based on statistical testing is performed to address dimensionality problem. With a total of 40 participants, this approach has provided encouraging Fl-Scores.