Section: New Results

Pathological voice classification

Participants : Khalid Daoudi, Nicolas Brodu.

Based on our GCI detection algorithm, we redefined the classical pitch perturbation measures that are widely used in voice quality assessment. We showed that our perturbation measures yield significantly better performance in pathological voice classification than classical measures. We also showed that some matching pursuit features can allow good performances in discrimination between pathological voice categories.

Publications: [31] , [30] .