Comparative Analysis of Glottal and Vocal Tract Features in Dysarthria

Dysarthria is a neurological disorder associated with the muscles in the vocal tract system that is caused by problems with coordination. It affects the vocal tract system and the glottis with different levels of severity. This research deals with the role of vocal tract (i.e., filter) and glottal excitation source in dysarthric speech, using Glottal Flow Model with Iterative Adaptive Inverse Filtering (GFM-IAIF). By decomposing the vocal tract and glottal source into two separate components, we were able to identify which component (source or filter) is affected the most. Our research showed that the vocal tract system is the most affected part, determining a better classification of features the magnitude spectrum-based Mel frequency cepstral coefficient (MFCC) resulted in 95.75% accuracy for vocal tract components and lower accuracy of 86.5% for glottal source components. In the same way, Modified Group Delay Cepstral Coefficients (MGDCC) correlated with a test accuracy of 94.43% of the vocal tract in comparison with 88.70% of the glottal source. These outcomes reveal that the vocal tract gets most damaged due to dysarthria and thus, emphasis the use of specific diagnostic and therapy interventions targeted at this area. The research not only explores feature fusion but also recommends further efforts to refine dysarthria diagnosis and treatment.